practical programming - an introduction to computer science using python
TRANSCRIPT
Prepared exclusively for Trieu Nguyen
What Readers Are Saying About
Practical Programming
Practical Programming is true to its name. The information it presents
is organized around useful tasks rather than abstract constructs, and
each chapter addresses a well-contained and important aspect of pro-
gramming in Python. A student wondering “How do I make the com-
puter do X?” would be able to find their answer very quickly with this
book.
Christine Alvarado
Associate professor of computer science, Harvey Mudd College
Science is about learning by performing experiments. This book
encourages computer science students to experiment with short,
interactive Python scripts and in the process learn fundamental con-
cepts such as data structures, sorting and searching algorithms,
object-oriented programming, accessing databases, graphical user
interfaces, and good program design. Clearly written text along with
numerous compelling examples, diagrams, and images make this an
excellent book for the beginning programmer.
Ronald Mak
Research staff member, IBM Almaden Research Center
Lecturer, Department of Computer Science, San Jose State
University
What, no compiler, no sample payroll application? What kind of pro-
gramming book is this? A great one, that’s what. It launches from a
“You don’t know anything yet” premise into a fearless romp through
the concepts and techniques of relevant programming technology.
And what fun students will have with the images and graphics in the
exercises!
Laura Wingerd
Author, Practical Perforce
Prepared exclusively for Trieu Nguyen
The debugging section is truly excellent. I know several practicing pro-
grammers who’d be rightfully offended by a suggestion to study the
whole book but who could really do with brushing up on this section
(and many others) once in a while.
Alex Martelli
Author, Python in a Nutshell
This book succeeds in two different ways. It is both a science-focused
CS1 text and a targeted Python reference. Even as it builds students’
computational insights, it also empowers and encourages them to
immediately apply their newfound programming skills in the lab or
on projects of their own.
Zachary Dodds
Associate professor of computer science, Harvey Mudd College
Prepared exclusively for Trieu Nguyen
Prepared exclusively for Trieu Nguyen
Practical ProgrammingAn Introduction to Computer Science Using Python
Jennifer Campbell
Paul Gries
Jason Montojo
Greg Wilson
The Pragmatic BookshelfRaleigh, North Carolina Dallas, Texas
Prepared exclusively for Trieu Nguyen
Many of the designations used by manufacturers and sellers to distinguish their prod-
ucts are claimed as trademarks. Where those designations appear in this book, and The
Pragmatic Programmers, LLC was aware of a trademark claim, the designations have
been printed in initial capital letters or in all capitals. The Pragmatic Starter Kit, The
Pragmatic Programmer, Pragmatic Programming, Pragmatic Bookshelf and the linking g
device are trademarks of The Pragmatic Programmers, LLC.
Every precaution was taken in the preparation of this book. However, the publisher
assumes no responsibility for errors or omissions, or for damages that may result from
the use of information (including program listings) contained herein.
Our Pragmatic courses, workshops, and other products can help you and your team
create better software and have more fun. For more information, as well as the latest
Pragmatic titles, please visit us at
http://www.pragprog.com
Copyright © 2009 Jennifer Campbell, Paul Gries, Jason Montojo, and Greg Wilson.
All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system, or transmit-
ted, in any form, or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior consent of the publisher.
Printed in the United States of America.
ISBN-10: 1-934356-27-1
ISBN-13: 978-1-934356-27-2
Printed on acid-free paper.
P1.0 printing, April 2009
Version: 2009-5-6
Prepared exclusively for Trieu Nguyen
Contents1 Introduction 11
1.1 Programs and Programming . . . . . . . . . . . . . . . . 13
1.2 A Few Definitions . . . . . . . . . . . . . . . . . . . . . . 14
1.3 What to Install . . . . . . . . . . . . . . . . . . . . . . . . 14
1.4 For Instructors . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2 Hello, Python 17
2.1 The Big Picture . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Expressions . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 What Is a Type? . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Variables and the Assignment Statement . . . . . . . . 25
2.5 When Things Go Wrong . . . . . . . . . . . . . . . . . . 29
2.6 Function Basics . . . . . . . . . . . . . . . . . . . . . . . 30
2.7 Built-in Functions . . . . . . . . . . . . . . . . . . . . . . 33
2.8 Style Notes . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3 Strings 39
3.1 Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Escape Characters . . . . . . . . . . . . . . . . . . . . . 42
3.3 Multiline Strings . . . . . . . . . . . . . . . . . . . . . . . 43
3.4 Print . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Formatted Printing . . . . . . . . . . . . . . . . . . . . . 45
3.6 User Input . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Prepared exclusively for Trieu Nguyen
CONTENTS 8
4 Modules 50
4.1 Importing Modules . . . . . . . . . . . . . . . . . . . . . 50
4.2 Defining Your Own Modules . . . . . . . . . . . . . . . . 54
4.3 Objects and Methods . . . . . . . . . . . . . . . . . . . . 60
4.4 Pixels and Colors . . . . . . . . . . . . . . . . . . . . . . 68
4.5 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.6 Style Notes . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5 Lists 81
5.1 Lists and Indices . . . . . . . . . . . . . . . . . . . . . . 81
5.2 Modifying Lists . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3 Built-in Functions on Lists . . . . . . . . . . . . . . . . 87
5.4 Processing List Items . . . . . . . . . . . . . . . . . . . . 89
5.5 Slicing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.6 Aliasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.7 List Methods . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.8 Nested Lists . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.9 Other Kinds of Sequences . . . . . . . . . . . . . . . . . 99
5.10 Files as Lists . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.11 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.12 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.13 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6 Making Choices 108
6.1 Boolean Logic . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2 if Statements . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.3 Storing Conditionals . . . . . . . . . . . . . . . . . . . . 125
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7 Repetition 131
7.1 Counted Loops . . . . . . . . . . . . . . . . . . . . . . . . 131
7.2 while Loops . . . . . . . . . . . . . . . . . . . . . . . . . . 140
7.3 User Input Loops . . . . . . . . . . . . . . . . . . . . . . 148
7.4 Controlling Loops . . . . . . . . . . . . . . . . . . . . . . 149
7.5 Style Notes . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CONTENTS 9
8 File Processing 159
8.1 One Record per Line . . . . . . . . . . . . . . . . . . . . 160
8.2 Records with Multiple Fields . . . . . . . . . . . . . . . . 171
8.3 Positional Data . . . . . . . . . . . . . . . . . . . . . . . 174
8.4 Multiline Records . . . . . . . . . . . . . . . . . . . . . . 177
8.5 Looking Ahead . . . . . . . . . . . . . . . . . . . . . . . . 179
8.6 Writing to Files . . . . . . . . . . . . . . . . . . . . . . . . 181
8.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9 Sets and Dictionaries 185
9.1 Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
9.2 Dictionaries . . . . . . . . . . . . . . . . . . . . . . . . . 190
9.3 Inverting a Dictionary . . . . . . . . . . . . . . . . . . . . 197
9.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
9.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
10 Algorithms 203
10.1 Searching . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
10.2 Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
10.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
10.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
11 Searching and Sorting 214
11.1 Linear Search . . . . . . . . . . . . . . . . . . . . . . . . 214
11.2 Binary Search . . . . . . . . . . . . . . . . . . . . . . . . 218
11.3 Sorting . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
11.4 More Efficient Sorting Algorithms . . . . . . . . . . . . . 228
11.5 Mergesort: An N log2N Algorithm . . . . . . . . . . . . . 229
11.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
11.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
12 Construction 237
12.1 More on Functions . . . . . . . . . . . . . . . . . . . . . 237
12.2 Exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . 242
12.3 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
12.4 Debugging . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.5 Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
12.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CONTENTS 10
13 Object-Oriented Programming 270
13.1 Class Color . . . . . . . . . . . . . . . . . . . . . . . . . . 271
13.2 Special Methods . . . . . . . . . . . . . . . . . . . . . . . 276
13.3 More About dir and help . . . . . . . . . . . . . . . . . . 278
13.4 A Little Bit of OO Theory . . . . . . . . . . . . . . . . . . 280
13.5 A Longer Example . . . . . . . . . . . . . . . . . . . . . . 288
13.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
13.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
14 Graphical User Interfaces 294
14.1 The Tkinter Module . . . . . . . . . . . . . . . . . . . . . 295
14.2 Basic GUI Construction . . . . . . . . . . . . . . . . . . 296
14.3 Models, Views, and Controllers . . . . . . . . . . . . . . 301
14.4 Style . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
14.5 A Few More Widgets . . . . . . . . . . . . . . . . . . . . . 312
14.6 Object-Oriented GUIs . . . . . . . . . . . . . . . . . . . . 316
14.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
14.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
15 Databases 321
15.1 The Big Picture . . . . . . . . . . . . . . . . . . . . . . . 321
15.2 First Steps . . . . . . . . . . . . . . . . . . . . . . . . . . 323
15.3 Retrieving Data . . . . . . . . . . . . . . . . . . . . . . . 327
15.4 Updating and Deleting . . . . . . . . . . . . . . . . . . . 330
15.5 Transactions . . . . . . . . . . . . . . . . . . . . . . . . . 331
15.6 Using NULL for Missing Data . . . . . . . . . . . . . . . 333
15.7 Using Joins to Combine Tables . . . . . . . . . . . . . . 334
15.8 Keys and Constraints . . . . . . . . . . . . . . . . . . . . 339
15.9 Advanced Features . . . . . . . . . . . . . . . . . . . . . 341
15.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
15.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
A Bibliography 351
Index 353
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 1
IntroductionTake a look at the pictures in Figure 1.1, on the following page. The
first one shows forest cover in the Amazon basin in 1975. The second
one shows the same area 26 years later. Anyone can see that much of
the rainforest has been destroyed, but how much is “much”?
Now look at Figure 1.2, on page 13.
Are these blood cells healthy? Do any of them show signs of leukemia?
It would take an expert doctor a few minutes to tell. Multiply those
minutes by the number of people who need to be screened. There simply
aren’t enough human doctors in the world to check everyone.
This is where computers come in. Computer programs can measure the
differences between two pictures and count the number of oddly shaped
platelets in a blood sample. Geneticists use programs to analyze gene
sequences; statisticians, to analyze the spread of diseases; geologists, to
predict the effects of earthquakes; economists, to analyze fluctuations
in the stock market; and climatologists, to study global warming. More
and more scientists are writing programs to help them do their work. In
turn, those programs are making entirely new kinds of science possible.
Of course, computers are good for a lot more than just science. We used
computers to write this book; you have probably used one today to chat
with friends, find out where your lectures are, or look for a restaurant
that serves pizza and Chinese food. Every day, someone figures out how
to make a computer do something that has never been done before.
Together, those “somethings” are changing the world.
This book will teach you how to make computers do what you want
them to do. You may be planning to be a doctor, linguist, or physicist
Prepared exclusively for Trieu Nguyen
CHAPTER 1. INTRODUCTION 12
Figure 1.1: The Rainforest Retreats (Photo credit: NASA/Goddard Space
Flight Center Scientific Visualization Studio)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PROGRAMS AND PROGRAMMING 13
Figure 1.2: Healthy blood cells—or are they? (Photo credit: CDC)
rather than a full-time programmer, but whatever you do, being able
to program is as important as being able to write a letter or do basic
arithmetic.
We begin in this chapter by explaining what programs and program-
ming are. We then define a few terms and present a few boring-but-
necessary bits of information for course instructors.
1.1 Programs and Programming
A program is a set of instructions. When you write down directions to
your house for a friend, you are writing a program. Your friend “exe-
cutes” that program by following each instruction in turn.
Every program is written in terms of a few basic operations that its
reader already understands. For example, the set of operations that
your friend can understand might include the following: “Turn left at
Darwin Street,” “Go forward three blocks,” and “If you get to the gas
station, turn around—you’ve gone too far.”
Computers are similar but have a different set of operations. Some
operations are mathematical, like “Add 10 to a number and take the
square root,” while others include “Read a line from the file named
data.txt,” “Make a pixel blue,” or “Send email to the authors of this
book.”
The most important difference between a computer and an old-
fashioned calculator is that you can “teach” a computer new opera-
tions by defining them in terms of old ones. For example, you can teach
the computer that “Take the average” means “Add up the numbers in
a set and divide by the set’s size.” You can then use the operations you
have just defined to create still more operations, each layered on top of
the ones that came before. It’s a lot like creating life by putting atoms
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A FEW DEFINITIONS 14
together to make proteins and then combining proteins to build cells
and giraffes.
Defining new operations, and combining them to do useful things, is
the heart and soul of programming. It is also a tremendously powerful
way to think about other kinds of problems. As Prof. Jeannette Wing
wrote [Win06], computational thinking is about the following:
• Conceptualizing, not programming. Computer science is not com-
puter programming. Thinking like a computer scientist means
more than being able to program a computer. It requires think-
ing at multiple levels of abstraction.
• A way that humans, not computers, think. Computational thinking
is a way humans solve problems; it is not trying to get humans
to think like computers. Computers are dull and boring; humans
are clever and imaginative. We humans make computers exciting.
Equipped with computing devices, we use our cleverness to tackle
problems we would not dare take on before the age of computing
and build systems with functionality limited only by our imagina-
tions.
• For everyone, everywhere. Computational thinking will be a reality
when it is so integral to human endeavors it disappears as an
explicit philosophy.
We hope that by the time you have finished reading this book, you will
see the world in a slightly different way.
1.2 A Few Definitions
One of the pieces of terminology that causes confusion is what to call
certain characters. The Python style guide (and several dictionaries) use
these names, so this book does too:
() Parentheses
[ ] Brackets
{} Braces
1.3 What to Install
For current installation instructions, please download the code from
the book website and open install/index.html in a browser. The book URL
is http://pragprog.com/titles/gwpy/practical-programming.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FOR INSTRUCTORS 15
1.4 For Instructors
This book uses the Python programming language to introduce stan-
dard CS1 topics and a handful of useful applications. We chose Python
for several reasons:
• It is free and well documented. In fact, Python is one of the largest
and best-organized open source projects going.
• It runs everywhere. The reference implementation, written in C, is
used on everything from cell phones to supercomputers, and it’s
supported by professional-quality installers for Windows, Mac OS,
and Linux.
• It has a clean syntax. Yes, every language makes this claim, but in
the four years we have been using it at the University of Toronto,
we have found that students make noticeably fewer “punctuation”
mistakes with Python than with C-like languages.
• It is relevant. Thousands of companies use it every day; it is one of
the three “official languages” at Google, and large portions of the
game Civilization IV are written in Python. It is also widely used
by academic research groups.
• It is well supported by tools. Legacy editors like Vi and Emacs all
have Python editing modes, and several professional-quality IDEs
are available. (We use a free-for-students version of one called
Wing IDE.)
We use an “objects first, classes second” approach: students are shown
how to use objects from the standard library early on but do not create
their own classes until after they have learned about flow control and
basic data structures. This compromise avoids the problem of explain-
ing Java’s public static void main(String[ ] args) to someone who has never
programmed.
We have organized the book into two parts. The first covers fundamen-
tal programming ideas: elementary data types (numbers, strings, lists,
sets, and dictionaries), modules, control flow, functions, testing, debug-
ging, and algorithms. Depending on the audience, this material can be
covered in nine or ten weeks.
The second part of the book consists of more or less independent chap-
ters on more advanced topics that assume all the basic material has
been covered. The first of these chapters shows students how to create
their own classes and introduces encapsulation, inheritance, and poly-
morphism; courses for computer science majors will want to include
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 16
this material. The other chapters cover application areas, such as 3D
graphics, databases, GUI construction, and the basics of web program-
ming; these will appeal to both computer science majors and students
from the sciences and will allow the book to be used for both.
Lots of other good books on Python programming exist. Some are acces-
sible to novices [Guz04, Zel03], and others are for anyone with any
previous programming experience [DEM02, GL07, LA03]. You may also
want to take a look at [Pyt], the special interest group for educators
using Python.
1.5 Summary
In this book, we’ll do the following:
• We will show you how to develop and use programs that solve real-
world problems. Most of its examples will come from science and
engineering, but the ideas can be applied to any domain.
• We start by teaching you the core features of a programming lan-
guage called Python. These features are included in every modern
programming language, so you can use what you learn no matter
what you work on next.
• We will also teach you how to think methodically about program-
ming. In particular, we will show you how to break complex prob-
lems into simple ones and how to combine the solutions to those
simpler problems to create complete applications.
• Finally, we will introduce some tools that will help make your pro-
gramming more productive, as well as some others that will help
your applications cope with larger problems.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 2
Hello, PythonPrograms are made up of commands that a computer can understand.
These commands are called statements, which the computer executes.
This chapter describes the simplest of Python’s statements and shows
how they can be used to do basic arithmetic. It isn’t very exciting in its
own right, but it’s the basis of almost everything that follows.
2.1 The Big Picture
In order to understand what happens when you’re programming, you
need to have a basic understanding of how a program gets executed on
a computer. The computer itself is assembled from pieces of hardware,
including a processor that can execute instructions and do arithmetic,
a place to store data such as a hard drive, and various other pieces such
as computer monitor, a keyboard, a card for connecting to a network,
and so on.
To deal with all these pieces, every computer runs some kind of operat-
ing system, such as Microsoft Windows, Linux, or Mac OS X. An oper-
ating system, or OS, is a program; what makes it special is that it’s the
only program on the computer that’s allowed direct access to the hard-
ware. When any other program on the computer wants to draw on the
screen, find out what key was just pressed on the keyboard, or fetch
data from the hard drive, it sends a request to the OS (see Figure 2.1,
on the following page).
This may seem a roundabout way of doing things, but it means that
only the people writing the OS have to worry about the differences
between one network card and another. Everyone else—everyone ana-
lyzing scientific data or creating 3D virtual chat rooms—only has to
Prepared exclusively for Trieu Nguyen
THE BIG PICTURE 18
Hard Drive Monitor
Operating System
User Program
Figure 2.1: Talking to the operating system
learn their way around the OS, and their programs will then run on
thousands of different kinds of hardware.
Twenty-five years ago, that’s how most programmers worked. Today,
though, it’s common to add another layer between the programmer and
the computer’s hardware. When you write a program in Python, Java,
or Visual Basic, it doesn’t run directly on top of the OS. Instead, another
program, called an interpreter or virtual machine, takes your program
and runs it for you, translating your commands into a language the OS
understands. It’s a lot easier, more secure, and more portable across
operating systems than writing programs directly on top of the OS.
But an interpreter alone isn’t enough; it needs some way to interact
with the world. One way to do this is to run a text-oriented program
called a shell that reads commands from the keyboard, does what they
ask, and shows their output as text, all in one window. Shells exist for
various programming languages as well as for interacting with the OS;
we will be exploring Python in this chapter using a Python shell.
The more modern way to interact with Python is to use an integrated
development environment, or IDE. This is a full-blown graphical inter-
face with menus and windows, much like a web browser, word proces-
sor, or drawing program.
Our favorite IDE for student-sized programs is the free Wing 101, a
“lite” version of the professional tool.1
1. See http://www.wingware.com for details.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXPRESSIONS 19
Figure 2.2: A Python shell
Another fine IDE is IDLE, which comes bundled with Python. We prefer
Wing 101 because it was designed specifically for beginning program-
mers, but IDLE is a capable development environment.
The Wing 101 interface is shown in Figure 2.3, on the next page. The
top part is the editing pane where we will write Python programs; the
bottom half, labeled as Python Shell, is where we will experiment with
snippets of Python programs. We’ll use the top pane more when we get
to Chapter 4, Modules, on page 50; for now we’ll stick to the shell.
The >>> part is called a prompt, because it prompts us to type some-
thing.
2.2 Expressions
As we learned at the beginning of the chapter, Python commands are
called statements. One kind of statement is an expression statement, or
expression for short. You’re familiar with mathematical expressions like
3 + 4 and 2 - 3 / 5; each expression is built out of values like 2 and 3 /
5 and operators like + and -, which combine their operands in different
ways.
Like any programming language, Python can evaluate basic mathemat-
ical expressions. For example, the following expression adds 4 and 13:
Download basic/addition.cmd
>>> 4 + 13
17
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXPRESSIONS 20
Figure 2.3: The Wing 101 interface
When an expression is evaluated, it produces a single result. In the
previous expression, 4 + 13 produced the result 17.
Type int
It’s not surprising that 4 + 13 is 17. However, computers do not always
play by the rules you learned in primary school. For example, look at
what happens when we divide 17 by 10:
Download basic/int_div.cmd
>>> 17 / 10
1
You would expect the result to be 1.7, but Python produces 1 instead.
This is because every value in Python has a particular type, and the
types of values determine how they behave when they’re combined.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXPRESSIONS 21
Division in Python 3.0
In the latest version of Python (Python 3.0), 5 / 2 is 2.5 rather than2. Python 3.0 is currently less widely used than its predecessors,so the examples in this book use the “classic” behavior.
In Python, an expression involving values of a certain type produces
a value of that same type. For example, 17 and 10 are integers—in
Python, we say they are of type int. When we divide one by the other,
the result is also an int.
Notice that Python doesn’t round integer expressions. If it did, the
result would have been 2. Instead, it takes the floor of the interme-
diate result. If you want the leftovers, you can use Python’s modulo
operator (%) to return the remainder:
Download basic/int_mod.cmd
>>> 17 % 10
7
Be careful about using % and / with negative operands. Since Python
takes the floor of the result of an integer division, the result is one
smaller than you might expect:
Download basic/neg_int_div.cmd
>>> -17 / 10
-2
When using modulo, the sign of the result matches the sign of the
second operand:
Download basic/neg_int_mod.cmd
>>> -17 % 10
3
>>> 17 % -10
-3
Type float
Python has another type called float to represent numbers with frac-
tional parts. The word float is short for floating point, which refers to
the decimal point that moves around between digits of the number.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHAT Is A TYPE? 22
An expression involving two floats produces a float:
Download basic/float_div_intro.cmd
>>> 17.0 / 10.0
1.7
When an expression’s operands are an int and a float, Python automati-
cally converts the int to a float. This is why the following two expressions
both return the same answer as the earlier one:
Download basic/float_division.cmd
>>> 17.0 / 10
1.7
>>> 17 / 10.0
1.7
If you want, you can omit the zero after the decimal point when writing
a floating-point number:
Download basic/float_division2.cmd
>>> 17 / 10.
1.7
>>> 17. / 10
1.7
However, most people think this is bad style, since it makes your pro-
grams harder to read: it’s very easy to miss a dot on the screen and see
“17” instead of “17.”
2.3 What Is a Type?
We’ve now seen two types of numbers, so we ought to explain exactly
what we mean by a type. In computing, a type is a set of values, along
with a set of operations that can be performed on those values. For
example, the type int is the values ..., -3, -2, -1, 0, 1, 2, 3, ..., along with
the operators +, -, *, /, and % (and a few others we haven’t introduced
yet). On the other hand, 84.2 is a member of the set of float values, but
it is not in the set of int values.
Arithmetic was invented before Python, so the int and float types have
exactly the same operators. We can see what happens when these are
applied to various values in Figure 2.4, on the next page.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHAT Is A TYPE? 23
Operator Symbol Example Result
- Negation -5 -5
* Multiplication 8.5 * 2.5 29.75
/ Division 11 / 3 3
% Remainder 8.5 % 3.5 1.5
+ Addition 11 + 3 14
- Subtraction {5 - 19} -14
** Exponentiation 2 ** 5 32
Figure 2.4: Arithmetic operators
Finite Precision
Floating-point numbers are not exactly the fractions you learned in
grade school. For example, take a look at Python’s version of the frac-
tion 13 (remember to include a decimal point so that the result isn’t
truncated):
Download basic/rate.cmd
>>> 1.0 / 3.0
0.33333333333333331
What’s that 1 doing at the end? Shouldn’t it be a 3? The problem is
that real computers have a finite amount of memory, which limits how
much information they can store about any single number. The number
0.33333333333333331 turns out to be the closest value to 13 that the
computer can actually store.
Operator Precedence
Let’s put our knowledge of ints and floats to use to convert Fahrenheit to
Celsius. To do this, we subtract 32 from the temperature in Fahrenheit
and then multiply by 59 :
Download basic/precedence.cmd
>>> 212 - 32.0 * 5.0 / 9.0
194.22222222222223
Python claims the result is 194.222222222222232 degrees Celsius
when in fact it should be 100. The problem is that * and / have higher
2. This is another floating-point approximation.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHAT Is A TYPE? 24
More on Numeric Precision
Computers use the same amount of memory to store an inte-ger regardless of that integer’s value, which means that -22984,-1, and 100000000 all take up the same amount of room.Because of this, computers can store int values only in a certainrange. A modern desktop or laptop machine, for example, canstore the numbers only from -2147483648 to 2147483647. (We’lltake a closer look in the exercises at where these bounds comefrom.)
Computers can store only approximations to real numbers forthe same reason. For example, 1
4 can be stored exactly, but aswe’ve already seen, 1
3 cannot. Using more memory won’t solvethe problem, though it will make the approximation closer tothe real value, just as writing a larger number of 3s after the 0 in0.333... doesn’t make it exactly equal to 1
3 .
The difference between 13 and 0.33333333333333331 may look
tiny. But if we use that value in a calculation, then the errormay get compounded. For example, if we add the float toitself, the result ends in ...6662; that is a slightly worse approxima-tion to 2
3 than 0.666.... As we do more calculations, the round-ing errors can get larger and larger, particularly if we’re mix-ing very large and very small numbers. For example, supposewe add 10,000,000,000 and 0.00000000001. The result ought tohave twenty zeroes between the first and last significant digit,but that’s too many for the computer to store, so the result is just10,000,000,000—it’s as if the addition never took place. Addinglots of small numbers to a large one can therefore have noeffect at all, which is not what a bank wants when it totals upthe values of its customers’ savings accounts.
It’s important to be aware of the floating-point issue so thatyour programs don’t bite you unexpectedly, but the solutions tothis problem are beyond the scope of this text. In fact, numeri-cal analysis, the study of algorithms to approximate continuousmathematics, is one of the largest subfields of computer sci-ence and mathematics.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
VARIABLES AND THE ASSIGNMENT STATEMENT 25
Operator Symbol
** Exponentiation
- Negation
*, /, % Multiplication, division, and remainder
+- Addition and subtraction
Figure 2.5: Arithmetic operators by precedence
precedence than -; in other words, when an expression contains a mix
of operators, * and / are evaluated before - and +. This means that what
we actually calculated was 212 - ((32.0 * 5.0) / 9.0).
We can alter the order of precedence by putting parentheses around
parts of the expression, just as we did in Mrs. Singh’s fourth-grade
class:
Download basic/precedence_diff.cmd
>>> (212 - 32.0) * 5.0 / 9.0
100.0
The order of precedence for arithmetic operators is listed in Figure 2.5.
It’s a good rule to parenthesize complicated expressions even when you
don’t need to, since it helps the eye read things like 1+1.7+3.2*4.4-16/3.
2.4 Variables and the Assignment Statement
Most handheld calculators3 have one or more memory buttons. These
store a value so that it can be used later. In Python, we can do this
with a variable, which is just a name that has a value associated with
it. Variables’ names can use letters, digits, and the underscore symbol.
For example, X, species5618, and degrees_celsius are all allowed, but 777
isn’t (it would be confused with a number), and neither is no-way! (it
contains punctuation).
You create a new variable simply by giving it a value:
Download basic/assignment.cmd
>>> degrees_celsius = 26.0
3. And cell phones, and wristwatches, and...
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
VARIABLES AND THE ASSIGNMENT STATEMENT 26
degrees_celsius 26.0
Figure 2.6: Memory model for a variable and its associated value
This statement is called an assignment statement; we say that degrees_
celsius is assigned the value 26.0. An assignment statement is executed
as follows:
1. Evaluate the expression on the right of the = sign.
2. Store that value with the variable on the left of the = sign.
In Figure 2.6, we can see the memory model for the result of the assign-
ment statement. It’s pretty simple, but we will see more complicated
memory models later.
Once a variable has been created, we can use its value in other cal-
culations. For example, we can calculate the difference between the
temperature stored in degrees_celsius and the boiling point of water like
this:
Download basic/variable.cmd
>>> 100 - degrees_celsius
74.0
Whenever the variable’s name is used in an expression, Python uses
the variable’s value in the calculation. This means that we can create
new variables from old ones:
Download basic/assignment2.cmd
>>> difference = 100 - degrees_celsius
Typing in the name of a variable on its own makes Python display its
value:
Download basic/variable2.cmd
>>> difference
74.0
What happened here is that we gave Python a very simple expression—
one that had no operators at all—so Python evaluated it and showed us
the result.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
VARIABLES AND THE ASSIGNMENT STATEMENT 27
It’s no more mysterious than asking Python what the value of 3 is:
Download basic/simplevalue.cmd
>>> 3
3
Variables are called variables because their values can change as the
program executes. For example, we can assign difference a new value:
Download basic/variable3.cmd
>>> difference = 100 - 15.5
>>> difference
84.5
This does not change the results of any calculations done with that
variable before its value was changed:
Download basic/variable4.cmd
>>> difference = 20
>>> double = 2 * difference
>>> double
40
>>> difference = 5
>>> double
40
As the memory models illustrate in Figure 2.7, on the following page,
once a value is associated with double, it stays associated until the pro-
gram explicitly overwrites it. Changes to other variables, like difference,
have no effect.
We can even use a variable on both sides of an assignment statement:
Download basic/variable5.cmd
>>> number = 3
>>> number
3
>>> number = 2 * number
>>> number
6
>>> number = number * number
>>> number
36
This wouldn’t make much sense in mathematics—a number cannot be
equal to twice its own value—but = in Python doesn’t mean “equals to.”
Instead, it means “assign a value to.”
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
VARIABLES AND THE ASSIGNMENT STATEMENT 28
difference 20
difference 20
double 40
difference 5
double 40
>>> difference = 20
>>> double = 2 * difference
>>> difference = 5
Figure 2.7: Changing a variable’s value
When a statement like number = 2 * number is evaluated, Python does the
following:
1. Gets the value currently associated with number
2. Multiplies it by 2 to create a new value
3. Assigns that value to number
Combined Operators
In the previous example, variable number appeared on both sides of
the assignment statement. This is so common that Python provides a
shorthand notation for this operation:
Download basic/variable6.cmd
>>> number = 100
>>> number -= 80
>>> number
20
Here is how a combined operator is evaluated:
1. Evaluate the expression to the right of the = sign.
2. Apply the operator attached to the = sign to the variable and the
result of the expression.
3. Assign the result to the variable to the left of the = sign.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHEN THINGS GO WRONG 29
Note that the operator is applied after the expression on the right is
evaluated:
Download basic/variable7.cmd
>>> d = 2
>>> d *= 3 + 4
>>> d
14
All the operators in Figure 2.5, on page 25, have shorthand versions.
For example, we can square a number by multiplying it by itself:
Download basic/variable8.cmd
>>> number = 10
>>> number *= number
>>> number
100
which is equivalent to this:
Download basic/variable9.cmd
>>> number = 10
>>> number = number * number
>>> number
100
2.5 When Things Go Wrong
We said earlier that variables are created by assigning them values.
What happens if we try to use a variable that hasn’t been created yet?
Download basic/undefined_var.cmd
>>> 3 + something
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'something' is not defined
This is pretty cryptic. In fact, Python’s error messages are one of its few
weaknesses from the point of view of novice programmers. The first two
lines aren’t much use right now, though they’ll be indispensable when
we start writing longer programs. The last line is the one that tells us
what went wrong: the name something wasn’t recognized.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FUNCTION BASICS 30
Here’s another error message you might sometimes see:
Download basic/syntax_error.cmd
>>> 2 +
File "<stdin>", line 1
2 +
^
SyntaxError: invalid syntax
The rules governing what is and isn’t legal in a programming language
(or any other language) are called its syntax. What this message is
telling us is that we violated Python’s syntax rules—in this case, by
asking it to add something to 2 but not telling it what to add.
2.6 Function Basics
Earlier in this chapter, we converted 80 degrees Fahrenheit to Celsius.
A mathematician would write this as f(t)= 59 (t-32), where t is the tem-
perature in Fahrenheit that we want to convert to Celsius. To find out
what 80 degrees Fahrenheit is in Celsius, we replace t with 80, which
gives us f (80) = 59 (80-32), or 262
3 .
We can write functions in Python, too. As in mathematics, they are used
to define common formulas. Here is the conversion function in Python:
Download basic/fahr_to_cel.cmd
>>> def to_celsius(t):
... return (t - 32.0) * 5.0 / 9.0
...
It has these major differences from its mathematical equivalent:
• A function definition is another kind of Python statement; it de-
fines a new name whose value can be rather complicated but is
still just a value.
• The keyword def is used to tell Python that we’re defining a new
function.
• We use a readable name like to_celsius for the function rather than
something like f whose meaning will be hard to remember an hour
later. (This isn’t actually a requirement, but it’s good style.)
• There is a colon instead of an equals sign.
• The actual formula for the function is defined on the next line. The
line is indented four spaces and marked with the keyword return.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FUNCTION BASICS 31
Python displays a triple-dot prompt automatically when you’re in the
middle of defining a new function; you do not type the dots any more
than you type the greater-than signs in the usual >>> prompt. If you’re
using a smart editor, like the one in Wing 101, it will automatically
indent the body of the function by the required amount. (This is another
reason to use Wing 101 instead of a basic text editor like Notepad or
Pico: it saves a lot of wear and tear on your spacebar and thumb.)
Here is what happens when we ask Python to evaluate to_celsius(80),
to_celsius(78.8), and to_celsius(10.4):
Download basic/fahr_to_cel_2.cmd
>>> to_celsius(80)
26.666666666666668
>>> to_celsius(78.8)
26.0
>>> to_celsius(10.4)
-12.0
Each of these three statements is called a function call, because we’re
calling up the function to do some work for us. We have to define a
function only once; we can call it any number of times.
The general form of a function definition is as follows:
def function_name(parameters):
block
As we’ve already seen, the def keyword tells Python that we’re defin-
ing a new function. The name of the function comes next, followed by
zero or more parameters in parentheses and a colon. A parameter is
a variable (like t in the function to_celsius) that is given a value when
the function is called. For example, 80 was assigned to t in the func-
tion call to_celsius(80), and then 78.8 in to_celsius(78.8), and then 10.4
in to_celsius(10.4). Those actual values are called the arguments to the
function.
What the function does is specified by the block of statements inside
it. to_celsius’s block consisted of just one statement, but as we’ll see
later, the blocks making up more complicated functions may be many
statements long.
Finally, the return statement has this general form:
return expression
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FUNCTION BASICS 32
def to_celsius(t): return (t - 32.0) * 5.0 / 9.0 to_celsius(80)
(rest of program)
1
2
3
4
Figure 2.8: Function control flow
and is executed as follows:
1. Evaluate the expression to the right of the keyword return.
2. Use that value as the result of the function.
It’s important to be clear on the difference between a function definition
and a function call. When a function is defined, Python records it but
doesn’t execute it. When the function is called, Python jumps to the
first line of that function and starts running it (see Figure 2.8). When
the function is finished, Python returns to the place where the function
was originally called.
Local Variables
Some computations are complex, and breaking them down into sepa-
rate steps can lead to clearer code. Here, we break down the evaluation
of the polynomial ax2 + bx + c into several steps:
Download basic/multi_statement_block.cmd
>>> def polynomial(a, b, c, x):
... first = a * x * x
... second = b * x
... third = c
... return first + second + third
...
>>> polynomial(2, 3, 4, 0.5)
6.0
>>> polynomial(2, 3, 4, 1.5)
13.0
Variables like first, second, and third that are created within a function
are called local variables. These variables exist only during function
execution; when the function finishes executing, the variables no longer
exist. This means that trying to access a local variable from outside the
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BUILT -IN FUNCTIONS 33
function is an error, just like trying to access a variable that has never
been defined:
Download basic/local_variable.cmd
>>> polynomial(2, 3, 4, 1.3)
11.280000000000001
>>> first
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'first' is not defined
>>> a
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'a' is not defined
As you can see from this example, a function’s parameters are also local
variables. When a function is called, Python assigns the argument val-
ues given in the call to the function’s parameters. As you might expect,
if a function is defined to take a certain number of parameters, it must
be passed the same number of arguments:4
Download basic/matching_args_params.cmd
>>> polynomial(1, 2, 3)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: polynomial() takes exactly 4 arguments (3 given)
The scope of a variable is the area of the program that can access it.
For example, the scope of a local variable runs from the line on which
it is first defined to the end of the function.
2.7 Built-in Functions
Python comes with many built-in functions that perform common oper-
ations. One example is abs, which produces the absolute value of a
number:
Download basic/abs.cmd
>>> abs(-9)
9
4. We’ll see later how to create functions that take any number of arguments.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE NOTES 34
Another is round, which rounds a floating-point number to the nearest
integer:
Download basic/round.cmd
>>> round(3.8)
4.0
>>> round(3.3)
3.0
>>> round(3.5)
4.0
Just like user-defined functions, Python’s built-in functions can take
more than one argument. For example, we can calculate 24 using the
power function pow:
Download basic/two_args.cmd
>>> pow(2, 4)
16
Some of the most useful built-in functions are ones that convert from
one type to another. The type names int and float can be used as if they
were functions:
Download basic/typeconvert.cmd
>>> int(34.6)
34
>>> float(21)
21.0
In this example, we see that when a floating-point number is converted
to an integer and truncated, not rounded.
2.8 Style Notes
Psychologists have discovered that people can keep track of only a
handful of things at any one time [Hoc04]. Since programs can get quite
complicated, it’s important that you choose names for your variables
that will help you remember what they’re for. X1, X2, and blah won’t
remind you of anything when you come back to look at your program
next week; use names like celsius, average, and final_result instead.
Other studies have shown that your brain automatically notices differ-
ences between things—in fact, there’s no way to stop it from doing this.
As a result, the more inconsistencies there are in a piece of text, the
longer it takes to read. (JuSt thInK a bout how long It w o u l d tAKE
you to rEa d this cHaPTer iF IT wAs fORmaTTeD like thIs.) It’s therefore
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 35
also important to use consistent names for variables. If you call some-
thing maximum in one place, don’t call it max_val in another; if you use
the name max_val, don’t also use the name maxVal, and so on.
These rules are so important that many programming teams require
members to follow a style guide for whatever language they’re using,
just as newspapers and book publishers specify how to capitalize head-
ings and whether to use a comma before the last item in a list. If you
search the Internet for programming style guide, you’ll discover links to
hundreds of examples.
You will also discover that lots of people have wasted many hours argu-
ing over what the “best” style for code is. Some of your classmates may
have strong opinions about this as well. If they do, ask them what data
they have to back up their beliefs, in other words, whether they know of
any field studies that prove that spaces after commas make programs
easier to read than no spaces. If they can’t cite any studies, pat them
on the back and send them on their deluded way.
2.9 Summary
In this chapter, we learned the following:
• An operating system is a program that manages your computer’s
hardware on behalf of other programs. An interpreter or virtual
machine is a program that sits on top of the operating system and
runs your programs for you. Building layers like this is the best
way we have found so far for constructing complicated systems.
• Programs are made up of statements. These can be simple expres-
sions (which are evaluated immediately), assignment statements
(which create new variables or change the values of existing vari-
ables), and function definitions (which teach Python how to do
new things).
• Every value in Python has a specific type, which determines what
operations can be applied to it. The two types used to represent
numbers are int and float.
• Expressions are evaluated in a particular order. However, you can
change that order by putting parentheses around subexpressions.
• Variables must be given values before they are used.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 36
• When a function is called, the values of its arguments are assigned
to its parameters, the statements inside the function are executed,
and a value is returned. The values assigned to the function’s
parameters, and the values of any local variables created inside
the function, are forgotten after the function returns.
• Python comes with predefined functions called built-ins.
2.10 Exercises
Here are some exercises for you to try on your own:
1. For each of the following expressions, what value will the expres-
sion give? Verify your answers by typing the expressions into
Python.
a) 9 - 3
b) 8 * 2.5
c) 9 / 2
d) 9 / -2
e) 9 % 2
f) 9 % -2
g) -9 % 2
h) 9 / -2.0
i) 4 + 3 * 5
j) (4 + 3) * 5
2. Unary minus negates a number. Unary plus exists as well; for
example, Python understands +5. If x has the value -17, what do
you think +x should do? Should it leave the sign of the number
alone? Should it act like absolute value, removing any negation?
Use the Python shell to find out its behavior.
3. a) Create a new variable temp, and assign it the value 24.
b) Convert the value in temp from Celsius to Fahrenheit by mul-
tiplying by 1.8 and adding 32; associate the resulting value
with temp. What is temp’s new value?
4. a) Create a new variable x, and assign it the value 10.5.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 37
b) Create a new variable y, and assign it the value 4.
c) Sum x and y, and associate the resulting value with x. What
are x and y’s new values?
5. Write a bullet list description of what happens when Python eval-
uates the statement x += x - x when x has the value 3.
6. The function name to_celsius is problematic: it doesn’t mention the
original unit, and it isn’t a verb phrase. (Many function names
are verb phrases because functions actively do things.) We also
assumed the original unit was Fahrenheit, but Kelvin is a tem-
perature scale too, and there are many others (see Section 6.5,
Exercises, on page 127 for a discussion of them).
We could use a longer name such as fahrenheit_to_celsius or even
convert_fahrenheit_to_celsius. We could abbreviate it as fahr_to_cel,
make it much shorter and use f2c, or even just use f. Write a para-
graph describing which name you think is best and why. Consider
ease of remembering, ease of typing, and readability. Don’t forget
to consider people whose first language isn’t English.
7. In the United States, a car’s fuel efficiency is measured in miles
per gallon. In the metric system, it is usually measured in liters
per 100 kilometers.
a) Write a function called convert_mileage that converts from
miles per gallon to liters per 100 kilometers.
b) Test that your functions returns the right values for 20 and
40 miles per gallon.
c) How did you figure out what the right value was? How closely
do the computer’s results match the ones you expected?
8. Explain the difference between a parameter and an argument.
9. a) Define a function called liters_needed that takes a value repre-
senting a distance in kilometers and a value representing gas
mileage for a vehicle and returns the amount of gas needed
in liters to travel that distance. Your definition should call the
function convert_mileage that you defined as part of a previous
exercise.
b) Verify that liters_needed(150, 30) returns 11.761938367442955
and liters_needed(100, 30) returns 7.84129224496197.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 38
c) When liters_needed is called with arguments 100 and 30, what
is the value of the argument to convert_mileage?
d) The function call liters_needed(100, 30) results in a call to con-
vert_mileage. Which of those two functions finishes executing
first?
10. We’ve seen built-in functions abs, round, pow, int, and float. Using
these functions, write expressions that do the following:
a) Calculate 3 to the power of 7.
b) Convert 34.7 to an integer by truncating.
c) Convert 34.7 to an integer by rounding.
d) Take the absolute value of -86, then convert it to a floating-
point number.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 3
StringsNumbers are fundamental to computing—in fact, crunching numbers
is what computers were invented to do—but there are many other kinds
of data in the world as well, such as addresses, pictures, and music.
Each of these can be represented as a data type, and knowing how to
manipulate those data types is a big part of being able to program. This
chapter introduces a non-numeric data type that represents text, such
as the words in this sentence or a strand of DNA. Along the way, we
will see how to make programs a little more interactive.
3.1 Strings
Computers may have been invented to do arithmetic, but these days,
most of them spend a lot of their time processing text. From desktop
chat programs to Google, computers create text, store it, search it, and
move it from one place to another.
In Python, a piece of text is represented as a string, which is a sequence
of characters (letters, numbers, and symbols). The simplest data type
for storing sequences of characters is str; it can store characters from
the Latin alphabet found on most North American keyboards. Another
data type called unicode can store strings containing any characters at
all, including Chinese ideograms, chemical symbols, and Klingon. We
will use the simpler type, str, in our examples.
Prepared exclusively for Trieu Nguyen
STRINGS 40
In Python, we indicate that a value is a string by putting either single
or double quotes around it:
Download strings/string.cmd
>>> 'Aristotle'
'Aristotle'
>>> "Isaac Newton"
'Isaac Newton'
The quotes must match:
Download strings/mismatched_quotes.cmd
>>> 'Charles Darwin"
File "<stdin>", line 1
'Charles Darwin"
^
SyntaxError: EOL while scanning single-quoted string
We can join two strings together by putting them side by side:
Download strings/concat.cmd
>>> 'Albert' 'Einstein'
'AlbertEinstein'
Notice that the words Albert and Einstein run together. If we want a space
between the words, then we can add a space either to the end of Albert
or to the beginning of Einstein:
Download strings/concat_space.cmd
>>> 'Albert ' 'Einstein'
'Albert Einstein'
>>> 'Albert' ' Einstein'
'Albert Einstein'
It’s almost always clearer to join strings with +. When + has two string
operands, then it is referred to as the concatenation operator:
Download strings/concat2.cmd
>>> 'Albert' + ' Einstein'
'Albert Einstein'
Since the + operator is used for both numeric addition and for string
concatenation, we call this an overloaded operator. It performs different
functions based on the type of operands that it is applied to.
The shortest string is the empty string, containing no characters at all.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STRINGS 41
As the following example shows, it’s the textual equivalent of 0—adding
it to another string has no effect:
Download strings/empty_string.cmd
>>> ''
''
>>> "Alan Turing" + ''
'Alan Turing'
>>> "" + 'Grace Hopper'
'Grace Hopper'
Here is an interesting question: can the + operator be applied to a string
and numeric value? If so, what function would be applied, addition or
concatenation? We’ll give it a try:
Download strings/concat3.cmd
>>> 'NH' + 3
Traceback (most recent call last):
File "<stdin>", line 1, in ?
TypeError: cannot concatenate 'str' and 'int' objects
>>> 9 + ' planets'
Traceback (most recent call last):
File "<stdin>", line 1, in ?
TypeError: unsupported operand type(s) for +: 'int' and 'str'
This is the second time Python has told us that we have a type error.
The first time, in Section 2.6, Local Variables, on page 32, the problem
was not passing the right number of parameters to a function. Here,
Python took exception to our attempts to add values of different data
types, because it doesn’t know which version of + we want: the one that
adds numbers or the one that concatenates strings.
In this case, it’s easy for a human being to see what the right answer
is. But what about this example?
Download strings/concat4.cmd
>>> '123' + 4
Should Python produce the string ’1234’ or the integer 127? The answer
is that it shouldn’t do either: if it guesses what we want, it’ll be wrong
at least some of the time, and we will have to try to track down the
problem without an error message to guide us.1
1. If you still aren’t convinced, consider this: in JavaScript (a language used for web
programming), ’7’+0 is the string ’70’, but ’7’-0 is 7.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ESCAPE CHARACTERS 42
If you want to put a number in the middle of a string, the easiest way is
to convert it via the built-in str function and then do the concatenation:
Download strings/concat4.cmd
>>> '123' + 4
The fact that Python will not combine strings and numbers using +
doesn’t mean that other operators can’t combine strings and integers.
In particular, we can repeat a string using the * operator, like this:
Download strings/repeat.cmd
>>> 'AT' * 5
'ATATATATAT'
>>> 4 * '-'
'----'
If the integer is less than or equals to zero, then this operator yields the
empty string (a string containing no characters):
Download strings/repeat2.cmd
>>> 'GC' * 0
''
>>> 'TATATATA' * -3
''
3.2 Escape Characters
Suppose you want to put a single quote inside a string. If you write it
directly, Python will complain:
Download strings/single_in_single.cmd
>>> 'that's not going to work'
File "<stdin>", line 1
'that's not going to work'
^
SyntaxError: invalid syntax
The problem is that when Python sees the second quote—the one that
you think of as being part of the string—it thinks the string is over. It
then doesn’t know what to do with all the stuff that comes after the
second quote.
One simple way to fix this is to use double quotes around the string:
Download strings/single_in_double.cmd
>>> "that's better"
"that's better"
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MULTILINE STRINGS 43
Escape Sequence Description
\n End of line
\\ Backslash
\’ Single quote
\" Double quote
\t Tab
Figure 3.1: Escape sequences
If you need to put a double quote in a string, you can use single quotes
around the string. But what if you want to put both kinds of quote in
one string? You could do this:
Download strings/adding_quotes.cmd
>>> 'She said, "That' + "'" + 's hard to read."'
Luckily, there’s a better way. If you type the previous expression into
Python, the result is as follows:
Download strings/adding_quotes_output.cmd
'She said, "That\'s hard to read."'
The combination of the backslash and the single quote is called an
escape sequence. The name comes from the fact that we’re “escaping”
from Python’s usual syntax rules for a moment. When Python sees a
backslash inside a string, it means that the next character represents
something special—in this case, a single quote, rather than the end of
the string. The backslash is called an escape character, since it signals
the start of an escape sequence.
As shown in Figure 3.1, Python recognizes several escape sequences.
In order to see how most are used, we will have to introduce two more
ideas: multiline strings and printing.
3.3 Multiline Strings
If you create a string using single or double quotes, the whole string
must fit onto a single line.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PRINT 44
Here’s what happens when you try to stretch a string across multiple
lines:
Download strings/multi1.cmd
>>> 'one
Traceback (most recent call last):
File "<string>", line 1, in <string>
Could not execute because an error occurred:
EOL while scanning single-quoted string: <string>, line 1, pos 4:
'one
EOL stands for “end of line,” so in this error report, Python is saying
that it reached the end of the line before it found the end of the string.
To span multiple lines, put three single quotes or three double quotes
around the string instead of one of each. The string can then span as
many lines as you want:
Download strings/multi2.cmd
>>> '''one
... two
... three'''
'one\ntwo\nthree'
Notice that the string Python creates contains a \n sequence every-
where our input started a new line. In reality, each of the three major
operating systems uses a different set of characters to indicate the end
of a line. This set of characters is called a newline. On Linux, a newline
is one ’\n’ character; on Mac OS X, it is one ’\r’; and on Windows, the
ends of lines are marked with both characters as ’\r\n’.
Python always uses a single \n to indicate a newline, even on operating
systems like Windows that do things other ways. This is called normal-
izing the string; Python does this so that you can write exactly the same
program no matter what kind of machine you’re running on.
3.4 Print
So far, we have been able to display the value of only one variable or
expression at a time. Real programs often want to display more infor-
mation, such as the values of multiple variable values. This can be done
using a print statement:
Download strings/print3.cmd
>>> print 1 + 1
2
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FORMATTED PRINTING 45
>>> print "The Latin 'oryctolagus cuniculus' means 'domestic rabbit'."
The Latin 'Oryctolagus cuniculus' means 'domestic rabbit'.
The first statement does what you’d expect from the numeric examples
we’ve seen previously, but the second does something slightly different
from previous string examples: it strips off the quotes around the string
and shows us the string’s contents, rather than its representation. This
example makes the difference between the two even clearer:
Download strings/print4.cmd
>>> print 'In 1859, Charles Darwin revolutionized biology'
In 1859, Charles Darwin revolutionized biology
>>> print 'and our understanding of ourselves'
and our understanding of ourselves
>>> print 'by publishing "On the Origin of Species".'
by publishing "On the Origin of Species".
And the following example shows that when Python prints a string, it
prints the values of any escape sequences in the string, rather than
their backslashed representations:
Download strings/print5.cmd
>>> print 'one\ttwo\nthree\tfour'
one two
three four
This example shows how the tab character \t can be used to lay values
out in columns. A print statement takes a comma-separated list of items
to print and displays them on a line of their own. If no values are given,
print simply displays a blank line. You can use any mix of types in the
list; Python always inserts a single space between each value:
Download strings/print_var.cmd
>>> area = 3.14159 * 5 * 5
>>> print "The area of the circle is", area, "sq cm."
The area of the circle is 78.539750 sq cm.
3.5 Formatted Printing
Sometimes, Python’s default printing rules aren’t what we want. In
these cases, we can specify the exact format we want for our output
by providing Python with a format string:
Download strings/print.cmd
>>> print "The area of the circle is %f sq cm." % area
The area of the circle is 78.539750 sq cm.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USER INPUT 46
In the previous statement, %f is a conversion specifier. It indicates where
the value of the variable area is to be inserted. Other markers that we
might use are %s, to insert a string value, and %d, to insert an integer.
The letter following the % is called the conversion type.
The % between the string and the value being inserted is another over-
loaded operator. We used % earlier for modulo; here, it is the string
formatting operator. It does not modify the string on its left side, any
more than the + in 3 + 5 changes the value of 3. Instead, the string
formatting operator returns a new string.
We can use the string formatting operator to lay out several values at
once. Here, for example, we are laying out a float and an int at the same
time:
Download strings/print2.cmd
>>> rabbits = 17
>>> cage = 10
>>> print "%f rabbits are in cage #%d." % (rabbits, cage)
17.000000 rabbits are in cage #10.
As we said earlier, print automatically puts a newline at the end of a
string. This isn’t necessarily what we want; for example, we might want
to print several pieces of data separately and have them all appear on
one line. To prevent the newline from being added, put a comma at the
end of the print statement:
Download strings/print_multiline2.cmd
>>> print rabbits,
17>>>
3.6 User Input
In an earlier chapter, we explored some built-in functions. Another
built-in function that you will find useful is raw_input, which reads a sin-
gle line of text from the keyboard. The “raw” part means that it returns
whatever the user enters as a string, even if it looks like a number:
Download strings/user_input.cmd
>>> line = raw_input()
Galapagos Islands
>>> print line
Galapagos Islands
>>> line = raw_input()
123
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 47
>>> print line * 2
123123
If you are expecting the user to enter a number, you must use int or
float to convert the string to the required type:
Download strings/user_input2.cmd
>>> value = raw_input()
123
>>> value = int(value)
>>> print value * 2
246
>>> value = float(raw_input())
Galapagos
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ValueError: invalid literal for float(): Galapagos
Finally, raw_input can be given a string argument, which is used to
prompt the user for input:
Download strings/raw_input_param.cmd
>>> name = raw_input("Please enter a name: ")
Please enter a name: Darwin
>>> print name
Darwin
3.7 Summary
In this chapter, we learned the following:
• Python uses the string type str to represent text as sequences of
characters.
• Strings are usually created by placing pairs of single or double
quotes around the text. Multiline strings can be created using
matching pairs of triple quotes.
• Special characters like newline and tab are represented using es-
cape sequences that begin with a backslash.
• Values can be displayed on the screen using a print statement and
input can be provided by the user using raw_input.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 48
3.8 Exercises
Here are some exercises for you to try on your own:
1. For each of the following expressions, what value will the expres-
sion give? Verify your answers by typing the expressions into the
Python shell.
a) ’Comp’ ’Sci’
b) ’Computer’ + ’ Science’
c) ’H20’ * 3
d) ’C02’ * 0
2. For each of the following phrases, express them as Python strings
using the appropriate type of quotation marks (single, double or
triple) and, if necessary, escape sequences:
a) They’ll hibernate during the winter.
b) “Absolutely not,” he said.
c) “He said, ’Absolutely not,”’ recalled Mel.
d) hydrogen sulfide
e) left\right
3. Rewrite the following string using single or double quotes instead
of triple quotes:
'''A
B
C'''
4. Use the built-in function len to find the length of the empty string.
5. Given variables x and y, which refer to values 3 and 12.5 respec-
tively, use print to display the following messages. When numbers
appear in the messages, the variables x and y should be used in
the print statement.
a) The rabbit is 3.
b) The rabbit is 3 years old.
c) 12.5 is average.
d) 12.5 * 3
e) 12.5 * 3 is 37.5.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 49
6. State whether each expression listed here evaluates to True or False:
a) ’g’ == "g"
b) ’g’ == ’G’
c) ’a’ ≥ ’b’
d) ’ant’ < ’abc’
e) ’ant’ > ’Ant’
f) ’ant’ > ’Abc’
g) ’ant’ < ’anti’
7. Use raw_input to prompt the user for a number and store the num-
ber entered as a float in a variable named num, and then print the
contents of num.
8. If you enter two strings side by side in Python, it automatically
concatenates them:
>>> 'abc' 'def'
'abcdef'
If those same strings are stored in variables, though, putting them
side by side is a syntax error:
>>> left = 'abc'
>>> right = 'def'
>>> left right
File "<stdin>", line 1
left right
^
SyntaxError: invalid syntax
Why do you think Python doesn’t let you do this?
9. Some people believe that multiplying a string by a negative num-
ber ought to produce an error, rather than an empty string. Ex-
plain why they might think this. If you agree, explain why; if you
don’t, explain why not.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 4
ModulesMathematicians don’t prove every theorem from scratch. Instead, they
build their proofs on the truths their predecessors have already estab-
lished. In the same way, it’s vanishingly rare for someone to write all of
a program herself; it’s much more common—and productive—to make
use of the millions of lines of code that other programmers have written
before.
A module is a collection of functions that are grouped together in a
single file. Functions in a module are usually related to each other in
some way; for example, the math module contains mathematical func-
tions such as cos (cosine) and sqrt (square root). This chapter shows you
how to use some of the hundreds of modules that come with Python and
how to create new modules of your own. You will also see how you can
use Python to explore and view images.
4.1 Importing Modules
When you want to refer to someone else’s work in a scientific paper, you
have to cite it in your bibliography. When you want to use a function
from a module, you have to import it. To tell Python that you want
to use functions in the math module, for example, you use this import
statement:
Download modules/import.cmd
>>> import math
Prepared exclusively for Trieu Nguyen
IMPORTING MODULES 51
Once you have imported a module, you can use the built-in help func-
tion to see what it contains:1
Download modules/help_math.cmd
>>> help(math)
Help on built-in module math:
NAME
math
FILE
(built-in)
DESCRIPTION
This module is always available. It provides access to the
mathematical functions defined by the C standard.
FUNCTIONS
acos(...)
acos(x)
Return the arc cosine (measured in radians) of x.
asin(...)
asin(x)
Return the arc sine (measured in radians) of x.
...
Great—our program can now use all the standard mathematical func-
tions. When we try to calculate a square root, though, we get an error
telling us that Python is still unable to find the function sqrt:
Download modules/sqrt.cmd
>>> sqrt(9)
Traceback (most recent call last):
File "<string>", line 1, in <string>
NameError: name 'sqrt' is not defined
The solution is to tell Python explicitly to look for the function in the
math module by combining the module’s name with the function’s name
using a dot:
Download modules/sqrt2.cmd
>>> math.sqrt(9)
3.0
1. When you do this interactively, Python displays only a screenful of information at a
time. Press the spacebar when you see the “More” prompt to go to the next page.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IMPORTING MODULES 52
floor
math
module
building
module
floor
floor(22.7)
?
Figure 4.1: How import works
The reason we have to join the function’s name with the module’s name
is that several modules might contain functions with the same name.
For example, does the following call to floor refer to the function from
the math module that rounds a number down or the function from the
(completely fictional) building module that calculates a price given an
area (see Figure 4.1)?
Download modules/import_ambiguity.cmd
>>> import math
>>> import building
>>> floor(22.7)
Once a module has been imported, it stays in memory until the program
ends. There are ways to “unimport” a module (in other words, to erase
it from memory) or to reimport a module that has changed while the
program is running, but they are rarely used. In practice, it’s almost
always simpler to stop the program and restart it.
Modules can contain more than just functions. The math module, for
example, also defines some variables like pi. Once the module has been
imported, you can use these variables like any others:
Download modules/pi.cmd
>>> math.pi
3.1415926535897931
>>> radius = 5
>>> print 'area is %6f' % (math.pi * radius ** 2)
area is 78.539816
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IMPORTING MODULES 53
You can even assign to variables imported from modules:
Download modules/pi_change.cmd
>>> import math
>>> math.pi = 3 # would turn circles into hexagons
>>> radius = 5
>>> print 'circumference is', 2 * math.pi * radius
circumference is 30
Don’t do this! Changing the value of π is not a good idea. In fact, it’s
such a bad idea that many languages allow programmers to define
unchangeable constants as well as variables. As the name suggests,
the value of a constant cannot be changed after it has been defined:
π is always 3.14159 and a little bit, while SECONDS_PER_DAY is always
86,400. The fact that Python doesn’t allow programmers to “freeze” val-
ues like this is one of the language’s few significant flaws.
Combining the module’s name with the names of the things it contains
is safe, but it isn’t always convenient. For this reason, Python lets you
specify exactly what you want to import from a module, like this:
Download modules/from.cmd
>>> from math import sqrt, pi
>>> sqrt(9)
3.0
>>> radius = 5
>>> print 'circumference is %6f' % (2 * pi * radius)
circumference is 31.415927
This can lead to problems when different modules provide functions
that have the same name. If you import a function called spell from a
module called magic and then you import another function called spell
from the module grammar, the second replaces the first. It’s exactly
like assigning one value to a variable, then another: the most recent
assignment or import wins.
This is why it’s usually not a good idea to use import *, which brings in
everything from the module at once. It saves some typing:
Download modules/from2.cmd
>>> from math import *>>> '%6f' % sqrt(8)
'2.828427'
but using it means that every time you add anything to a module, you
run the risk of breaking every program that uses it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEFINING YOUR OWN MODULES 54
The standard Python library contains several hundred modules to do
everything from figuring out what day of the week it is to fetching data
from a website. The full list is online at http://docs.python.org/modindex.
html; although it’s far too much to absorb in one sitting (or even one
course), knowing how to use the library well is one of the things that
distinguishes good programmers from poor ones.
4.2 Defining Your Own Modules
Section 2.1, The Big Picture, on page 17 explained that in order to save
code for later use, you can put it in a file with a .py extension. You
can then tell Python to run the code in that file, rather than typing
commands in at the interactive prompt. What we didn’t tell you then is
that every Python file can be used as a module. The name of the module
is the same as the name of the file, but without the .py extension.
For example, the following function is taken from Section 2.6, Function
Basics, on page 30:
Download modules/convert.py
def to_celsius(t):
return (t - 32.0) * 5.0 / 9.0
Put this function definition in a file called temperature.py, and then add
another function called above_freezing that returns True if its argument’s
value is above freezing (in Celsius), and False otherwise:
Download modules/freezing.py
def above_freezing(t):
return t > 0
Congratulations—you have now created a module called temperature:
Download modules/temperature.py
def to_celsius(t):
return (t - 32.0) * 5.0 / 9.0
def above_freezing(t):
return t > 0
Now that you’ve created this file, you can now import it like any other
module:
Download modules/import_temp.cmd
>>> import temperature
>>> temperature.above_freezing(temperature.to_celsius(33.3))
True
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEFINING YOUR OWN MODULES 55
The __builtins__ Module
Python’s built-in functions are actually in a module named__builtins__. The double underscores before and after the namesignal that it’s part of Python; we’ll see this convention usedagain later for other things. You can see what’s in the moduleusing help(__builtins__), or if you just want a directory, you can usedir instead (which works on other modules as well):
Download modules/dir1.cmd
>>> dir(__builtins__)['ArithmeticError', 'AssertionError', 'AttributeError','BaseException', 'DeprecationWarning', 'EOFError', 'Ellipsis','EnvironmentError', 'Exception', 'False', 'FloatingPointError','FutureWarning', 'GeneratorExit', 'IOError', 'ImportError','ImportWarning', 'IndentationError', 'IndexError', 'KeyError','KeyboardInterrupt', 'LookupError', 'MemoryError', 'NameError','None', 'NotImplemented', 'NotImplementedError', 'OSError','OverflowError', 'PendingDeprecationWarning', 'ReferenceError','RuntimeError', 'RuntimeWarning', 'StandardError','StopIteration', 'SyntaxError', 'SyntaxWarning', 'SystemError','SystemExit', 'TabError', 'True', 'TypeError','UnboundLocalError', 'UnicodeDecodeError', 'UnicodeEncodeError','UnicodeError', 'UnicodeTranslateError', 'UnicodeWarning','UserWarning', 'ValueError', 'Warning', 'ZeroDivisionError', '_','__debug__', '__doc__', '__import__', '__name__', 'abs', 'all','any', 'apply', 'basestring', 'bool', 'buffer', 'callable','chr', 'classmethod', 'cmp', 'coerce', 'compile', 'complex','copyright', 'credits', 'delattr', 'dict', 'dir', 'divmod','enumerate', 'eval', 'execfile', 'exit', 'file', 'filter','float', 'frozenset', 'getattr', 'globals', 'hasattr', 'hash','help', 'hex', 'id', 'input', 'int', 'intern', 'isinstance','issubclass', 'iter', 'len', 'license', 'list', 'locals', 'long','map', 'max', 'min', 'object', 'oct', 'open', 'ord', 'pow','property', 'quit', 'range', 'raw_input', 'reduce', 'reload','repr', 'reversed', 'round', 'set', 'setattr', 'slice', 'sorted','staticmethod', 'str', 'sum', 'super', 'tuple', 'type', 'unichr','unicode', 'vars', 'xrange', 'zip']
As of Python 2.5, 32 of the 135 things in __builtins__ are used tosignal errors of particular kinds, such as SyntaxError and ZeroDi-
visionError. There are also functions called copyright, which tellsyou who holds the copyright on Python, and license, which dis-plays Python’s rather complicated license. We’ll meet some ofthis module’s other members in later chapters.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEFINING YOUR OWN MODULES 56
Figure 4.2: The temperature module in Wing 101
What Happens During Import
Let’s try another experiment. Put the following in a file called experi-
ment.py:
Download modules/experiment.py
print "The panda's scientific name is 'Ailuropoda melanoleuca'"
and then import it (or click Wing 101’s Run button):
Download modules/import_experiment.cmd
>>> import experiment
The panda's scientific name is 'Ailuropoda melanoleuca'
What this shows is that Python executes modules as it imports them.
You can do anything in a module you would do in any other program,
because as far as Python is concerned, it’s just another bunch of state-
ments to be run.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEFINING YOUR OWN MODULES 57
Let’s try another experiment. Start a fresh Python session, and try
importing the experiment module twice in a row:
Download modules/import_twice.cmd
>>> import experiment
The panda's scientific name is 'Ailuropoda melanoleuca'
>>> import experiment
>>>
Notice that the message wasn’t printed the second time. That’s because
Python loads modules only the first time they are imported. Internally,
Python keeps track of the modules it has already seen; when it is asked
to load one that’s already in that list, it just skips over it. This saves time
and will be particularly important when you start writing modules that
import other modules, which in turn import other modules—if Python
didn’t keep track of what was already in memory, it could wind up
loading commonly used modules like math dozens of times.
Using __main__
As we’ve now seen, every Python file can be run directly from the com-
mand line or IDE or can be imported and used by another program. It’s
sometimes useful to be able to tell inside a module which is happening,
in other words, whether the module is the main program that the user
asked to execute or whether some other module has that honor.
Python defines a special variable called __name__ in every module to
help us figure this out. Suppose we put the following into echo.py:
Download modules/echo.py
print "echo: __name__ is", __name__
If we run this file, its output is as follows:
Download modules/echo.out
echo: __name__ is __main__
As promised, Python has created the variable __name__. Its value is
"__main__", meaning, “This module is the main program.”
But look at what happens when we import echo.py, instead of running
it directly:
Download modules/echo.cmd
>>> import echo
echo: __name__ is echo
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEFINING YOUR OWN MODULES 58
The same thing happens if we write a program that does nothing but
import our echoing module:
Download modules/import_echo.py
import echo
print "After import, __name__ is", __name__, "and echo.__name__ is", echo.__name__
which, when run from the command line, produces this:
Download modules/import_echo.out
echo: __name__ is echo
After import, __name__ is __main__ and echo.__name__ is echo
What’s happening here is that when Python imports a module, it sets
that module’s __name__ variable to be the name of the module, rather
than the special string "__main__". This means that a module can tell
whether it is the main program:
Download modules/test_main.py
if __name__ == "__main__":
print "I am the main program"
else:
print "Someone is importing me"
Try it, and see what happens when you run it directly and when you
import it.
Knowing whether a module is being imported or not turns out to allow a
few handy programming tricks. One is to provide help on the command
line whenever someone tries to run a module that’s meant to be used
as a library. For example, think about what happens when you run the
following on the command line vs. importing it into another program:
Download modules/main_help.py
'''
This module guesses whether something is a dinosaur or not.
'''
def is_dinosaur(name):
'''
Return True if the named create is recognized as a dinosaur,
and False otherwise.
'''
return name in ['Tyrannosaurus', 'Triceratops']
if __name__ == '__main__':
help(__name__)
We will see other uses in the following sections and in later chapters.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEFINING YOUR OWN MODULES 59
Providing Help
Let’s return to the temperature module for a moment and modify it to
round temperatures off. We’ll put the result in temp_round.py:
Download modules/temp_round.py
def to_celsius(t):
return round((t - 32.0) * 5.0 / 9.0)
def above_freezing(t):
return t > 0
What happens if we ask for help on the function to_celsius?
Download modules/help_temp.cmd
>>> import temp_round
>>> help(temp_round)
Help on module temp_round:
NAME
temp_round
FILE
/home/pybook/modules/temp_round.py
FUNCTIONS
above_freezing(t)
to_celsius(t)
That’s not much use: we know the names of the functions and how
many parameters they need, but not much else. To provide something
more useful, we should add docstrings to the module and the functions
it contains and save the result in temp_with_doc.py:
Download modules/temp_with_doc.py
'''Functions for working with temperatures.'''
def to_celsius(t):
'''Convert Fahrenheit to Celsius.'''
return round((t - 32.0) * 5.0 / 9.0)
def above_freezing(t):
'''True if temperature in Celsius is above freezing, False otherwise.'''
return t > 0
Asking for help on this module produces a much more useful result.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 60
Download modules/help_temp_with_doc.cmd
>>> import temp_with_doc
>>> help(temp_with_doc)
Help on module temp_with_doc:
NAME
temp_with_doc - Functions for working with temperatures.
FILE
/home/pybook/modules/temp_with_doc.py
FUNCTIONS
above_freezing(t)
True if temperature in Celsius is above freezing, False otherwise.
to_celsius(t)
Convert Fahrenheit to Celsius.
The term docstring is short for “documentation string.” Docstrings are
easy to create: if the first thing in a file or a function is a string that
isn’t assigned to anything, Python saves it so that help can print it later.
You might think that a module this small doesn’t need much documen-
tation. After all, it has only two functions, and their names are pretty
descriptive of what they do. But writing documentation is more than a
way to earn a few extra marks—it’s essential to making software usable.
Small programs have a way of turning into larger and more complicated
ones. If you don’t document as you go along and keep the documenta-
tion in the same file as the program itself, you will quickly lose track of
what does what.
4.3 Objects and Methods
Numbers and strings may have been enough to keep programmers
happy back in the twentieth century, but these days, people expect
to work with images, sound, and video as well. A Python module called
media provides functions for manipulating and viewing pictures; it isn’t
in the standard library, but it can be downloaded for free from http://
code.google.com/p/pygraphics/. (One of the exercises discusses why it
needs a separate download.)
In order to understand how media works, we first have to introduce two
concepts that are fundamental to modern program design. And to do
that, we have to back up and take another look at strings.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 61
So far, we have seen two operators that work on strings: concatena-
tion (+), which “adds” strings, and formatting (%), which gives you con-
trol over how values are displayed. There are dozens of other things
we might want to do to strings, such as capitalize them, strip off any
leading or trailing blanks, or find out whether one string is contained
inside another. Having single-character operators such as + and - for all
of these is impractical, because we would quickly run out of letters and
have to start using two- and three-character combinations that would
be impossible to remember.
We could put all the functions that work on strings in a module and
ask users to load that module, but there’s a simpler way to solve the
problem. Python strings “own” a set of special functions called methods.
These are called just like the functions inside a module. If we have a
string like ’hogwarts’, we can capitalize it by calling ’hogwarts’.capitalize(),
which returns ’Hogwarts’. Similarly, if the variable villain has been as-
signed the string ’malfoy’, the expression villain.capitalize() will return
the string ’Malfoy’.
Every string we create automatically shares all the methods that belong
to the string data type. The most commonly used ones are listed in
Figure 4.3, on the next page; you can find the complete list in Python’s
online documentation or type help(str) into the command prompt.
Using methods is almost the same as using functions, though a method
almost always does something to or with the thing that owns it. For
example, let’s call the startswith method on the string ’species’:
Download modules/startswith.cmd
>>> 'species'.startswith('a')
False
>>> 'species'.startswith('s')
True
The method startswith takes a string argument and returns a bool to tell
us whether the string whose method was called—the one on the left of
the dot—starts with the string that is given as an argument. String also
has an endswith method:
Download modules/endswith.cmd
>>> 'species'.endswith('a')
False
>>> 'species'.endswith('s')
True
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 62
Method Description
capitalize() Returns a copy of the string with the first letter cap-
italized
find(s) Returns the index of the first occurrence of s in the
string, or -1 if s is not in the string
find(s, beg) Returns the index of the first occurrence of s after
index beg in the string, or -1 if s is not in the string
after index beg
find(s, beg, end) Returns the index of the first occurrence of s between
indices beg and end in the string, or -1 if s is not in
the string between indices beg and end
islower() Tests that all characters are lowercase
isupper() Tests that all characters are uppercase
lower() Returns a copy of the string with all characters con-
verted to lowercase
replace(old, new) Returns a copy of the string with all occurrences of
the substring old replaced with new
split() Returns the space-separated words as a list
split(del) Returns the del-separated words as a list
strip() Returns a copy of the string with leading and trailing
whitespace removed
strip(s) Returns a copy of the string with the characters in s
removed
upper() Returns a copy of the string with all characters con-
verted to uppercase
Figure 4.3: Common string methods
We can chain multiple method calls together in a single line by calling
a method of the value returned by another method call. To show how
this works, let’s start by calling swapcase to change lowercase letters to
uppercase and uppercase to lowercase:
Download modules/swap.cmd
>>> 'Computer Science'.swapcase()
'cOMPUTER sCIENCE'
Since the result of this method is a string, we can immediately call the
result’s endswith method to check that the first call did the right thing
to the last few letters of the original string.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 63
'Computer Science'.swapcase().endswith('ENCE')
.endswith('ENCE')'cOMPUTER sCIENCE'
True
Figure 4.4: Chaining method calls
Download modules/swap_endswith.cmd
>>> 'Computer Science'.swapcase().endswith('ENCE')
True
In Figure 4.4, we can see what’s going on when we do this. Note that
Python automatically creates a temporary variable to hold the value of
the swapcase method call long enough for it to call that value’s endswith
method.
Something that has methods is called an object. It turns out that every-
thing in Python is an object, even the number zero:
Download modules/int_help.cmd
>>> help(0)
Help on int object:
class int(object)
| int(x[, base]) -> integer
|
| Convert a string or number to an integer, if possible. A floating point
| argument will be truncated towards zero (this does not include a string
| representation of a floating point number!) When converting a string, use
| the optional base. It is an error to supply a base when converting a
| non-string. If the argument is outside the integer range a long object
| will be returned instead.
|
| Methods defined here:
|
| __abs__(...)
| x.__abs__() <==> abs(x)
|
| __add__(...)
| x.__add__(y) <==> x+y
...
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 64
Most modern programming languages are structured this way: the
“things” in the program are objects, and most of the code in the pro-
gram consists of methods that use the data stored in those objects.
Chapter 13, Object-Oriented Programming, on page 270 will show you
how to create new kinds of objects; for now, let’s take a look at the
objects Python uses to store and manipulate images.
Images
Now that we have seen the basic features of modules, objects, and
methods, let’s look at how they can solve real-world problems. For our
running example, we will write some programs that display and manip-
ulate pictures and other images.
Suppose you have a file called pic207.jpg on your hard drive and want
to display it on your screen. You could double-click to open it, but
what does that actually do? To start to answer that question, type the
following into a Python prompt:
Download modules/open_pic.cmd
>>> import media
>>> f = media.choose_file()
>>> pic = media.load_picture(f)
>>> media.show(pic)
When the file dialog box opens, navigate to pic207.jpg. The result should
be the awesomely cute photo shown in Figure 4.5, on the following
page. Here’s what the commands shown earlier actually did:
1. Import the functions from the media module.
2. Call that module’s choose_file function to open a file-choosing dia-
log box. This call returns a string that contains the path to the
picture file.
3. Call the module’s load_picture function to read the contents of the
picture file into memory. This creates a Python object, which is
assigned to the variable pic.
4. Call that module’s show function, which launches another pro-
gram to display the picture. Python has to launch another pro-
gram because it can’t print the picture out at the command line.
Double-clicking would definitely have been easier.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 65
Figure 4.5: Madeleine
But let’s see your mouse do this:
Download modules/pic_props.cmd
>>> pic.get_width()
500
>>> pic.get_height()
375
>>> pic.title
'modules/pic207.jpg'
The first two commands tell us how wide and high the picture is in
pixels. The third tells us the path to the file containing the picture.
Now try this:
Download modules/pic_crop.cmd
>>> media.crop_picture(pic, 150, 50, 450, 300)
>>> media.show(pic)
>>> media.save_as(pic, 'pic207cropped.jpg')
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 66
Figure 4.6: Madeleine cropped
As you can guess from the name, crop crops the picture. The upper-
left corner is (150, 50), and the lower-right corner is (450, 300); the
resulting picture is shown in Figure 4.6.
The code also shows the new picture and then writes it to a new file.
This file is saved in the current working directory, which by default is the
directory in which the program is running. On our system this happens
to be ’/Users/pgries/’.
Now let’s put Madeleine’s name on her hat. To do that, we use picture’s
add_text function; the result is shown in Figure 4.7, on the following
page.
Download modules/pic_text.cmd
>>> media.add_text(pic, 115, 40, 'Madeleine', media.magenta)
>>> media.show(pic)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECTS AND METHODS 67
Figure 4.7: Madeleine named
Function choose_file is useful for writing interactive programs, but when
we know exactly which files we want or we want more than one file, it’s
often easier to skip that navigation step. As an example, let’s open up
all three pictures of Madeleine in a single program:
Download modules/show_madeleine.py
import media
pic1 = media.load_picture('pic207.jpg')
media.show(pic1)
pic2 = media.load_picture('pic207cropped.jpg')
media.show(pic2)
pic3 = media.load_picture('pic207named.jpg')
media.show(pic3)
Since we haven’t specified what directory to find the files in, the pro-
gram looks for them in the current working directory. If the program
can’t find them there, it reports an error.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PIXELS AND COLORS 68
Color Value
black Color(0, 0, 0)
white Color(255, 255, 255)
red Color(255, 0, 0)
green Color(0, 255, 0)
blue Color(0, 0, 255)
magenta Color(255, 0, 255)
yellow Color(255, 255, 0)
aqua Color(0, 255, 255)
pink Color(255, 192, 203)
purple Color(128, 0, 128)
Figure 4.8: Example color values
4.4 Pixels and Colors
Most people want to do a lot more to pictures than just display them
and crop them. If you do a lot of digital photography, you may want to
remove the “red-eye” caused by your camera flash. You might also want
to convert pictures to black and white for printing, highlight certain
objects, and so on.
To do these things, you must work with the individual pixels that make
up the image. The media module represents pixels using the RGB color
model discussed in the sidebar on page 72. Module media provides a
Color type and more than 100 predefined Color values. Several of them
are listed in Figure 4.3, on page 62; black is represented as “no blue,
no green, no red,” white is the maximum possible amount of all three,
and other colors lie somewhere in between.
The media module provides functions for getting and changing the col-
ors in pixels (see Figure 4.9, on the next page) and for manipulating
colors themselves (see Figure 4.10, on page 70).
To see how these functions are used, let’s go through all the pixels in
Madeleine’s cropped and named picture and make it look like it was
taken at sunset. To do this, we’re going to remove some of the blue
and some of the green from each pixel, making the picture darker and
redder.2
2. We’re not actually adding any red, but reducing the amount of blue and green will
fool the eye into thinking we have.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PIXELS AND COLORS 69
Function Description
get_red(pixel) Gets the red component of pixel
set_red(pixel, value) Sets the red component of pixel to value
get_blue(pixel) Gets the red component of pixel
set_blue(pixel, value) Sets the blue component of pixel to value
get_green(pixel) Gets the red component of pixel
set_green(pixel, value) Sets the green component of pixel to value
get_color(pixel) Gets the color of pixel
set_color(pixel, color) Sets the color of pixel to color
Figure 4.9: Pixel-manipulation functions
Download modules/sunset.py
import media
pic = media.load_picture('pic207.jpg')
media.show(pic)
for p in media.get_pixels(pic):
new_blue = int(0.7 * media.get_blue(p))
new_green = int(0.7 * media.get_green(p))
media.set_blue(p, new_blue)
media.set_green(p, new_green)
media.show(pic)
Some things to note:
• Color values are integers, so we need to convert the result of mul-
tiplying the blue and green by 0.7 using the function int.
• The for loop does something to each pixel in the picture. We will
talk about for loops in detail in Section 5.4, Processing List Items,
on page 89, but just reading the code aloud will give you the idea
that it associates each pixel in turn with the variable p, extracts
the blue and green components, calculates new values for them,
and then resets the values in the pixel.
Try this code on a picture of your own, and see how convincing the
result is.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 70
Function Description
darken(color) Returns a color slightly darker than color
lighten(color) Returns a color slightly darker than color
create_color(red, green, blue) Returns color (red, green, blue)
distance(c1, c2) Returns how far apart colors c1 and c2
are
Figure 4.10: Color functions
4.5 Testing
Another use for modules in real-world Python programming is to make
sure that programs don’t just run but also produce the right answers.
In science, for example, the programs you use to analyze experimen-
tal data must be at least as reliable as the lab equipment you used to
collect that data, or there’s no point running the experiment. The pro-
grams that run CAT scanners and other medical equipment must be
even more reliable, since lives depend on them. As it happens, the tools
used to make sure that these programs are behaving correctly can also
be used by instructors to grade students’ assignments and by students
to check their programs before submitting them.
Checking that software is doing the right thing is called quality assur-
ance, or QA. Over the last fifty years, programmers have learned that
quality isn’t some kind of magic pixie dust that you can sprinkle on a
program after it has been written. Quality has to be designed in, and
software must be tested and retested to check that it meets standards.
The good news is that putting effort into QA actually makes you more
productive overall. The reason can be seen in Boehm’s curve in Fig-
ure 4.11, on the following page. The later you find a bug, the more
expensive it is to fix, so catching bugs early reduces overall effort.
Most good programmers today don’t just test their software while writ-
ing it; they build their tests so that other people can rerun them months
later and a dozen time zones away. This takes a little more time up
front but makes programmers more productive overall, since every hour
invested in preventing bugs saves two, three, or ten frustrating hours
tracking bugs down.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 71
Requirements Design Coding Testing Deployment
Cost
Figure 4.11: Boehm’s curve
One popular testing library for Python is called Nose, which can be
downloaded for free at http://code.google.com/p/python-nose/. To show
how it works, we will use it to test our temperature module. To start,
create a new Python file called test_temperature.py. The name is impor-
tant: when Nose runs, it automatically looks for files whose names start
with the letters test_. The second part of the name is up to us—we could
call it test_hagrid.py if we wanted to—but a sensible name will make it
easier for other people to find things in our code.
Every Nose test module should contain the following:
• Statements to import Nose and the module to be tested
• Functions that actually test our module
• A function call to trigger execution of those test functions
Like the name of the test module, the names of the test functions must
start with test_. Using the structure outlined earlier, our first sketch of
a testing module looks like this:
Download modules/structure.py
import nose
import temperature
def test_to_celsius():
'''Test function for to_celsius'''
pass # we'll fill this in later
def test_above_freezing():
'''Test function for above_freezing.'''
pass # we'll fill this in too
if __name__ == '__main__':
nose.runmodule()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 72
RGB and Hexadecimal
In the red-green-blue (or RGB) color system, each pixel in a pic-ture has a certain amount of the three primary colors in it, andeach color component is specified by a number in the range0–255 (which is the range of numbers that can be representedin a single 8-bit byte).
By tradition, RGB values are represented in hexadecimal, orbase-16, rather than in the usual base-10 decimal system. The“digits” in hexadecimal are the usual 0–9, plus the letters A–F (ora–f). This means that the number after 916 is not 1016, but A16; thenumber after A16 is B16, and so on, up to F16, which is followedby 1016. Counting continues to 1F16, which is followed by 2016,and so on, up to FF16 (which is 1510×1610 + 1510, or 25510).
An RGB color is therefore six hexadecimal digits: two for red,two for green, and two for blue. Black is therefore #000000 (nocolor of any kind), while white is #FFFFFF (all colors saturated),and #008080 is a bluish-green (no red, half-strength green, half-strength blue).
For now, each test function contains nothing except a docstring and
a pass statement. As the name suggests, this does nothing—it’s just a
placeholder to remind ourselves that we need to come back and write
some more code.
If you run the test module, the output starts with two dots to say
that two tests have run successfully. (If a test fails, Nose prints an
“F” instead to attract attention to the problem.) The summary after the
dashed line tells us that Nose found and ran two tests, that it took less
than a millisecond to do so, and that everything was OK:
Download modules/structure.out
..
----------------------------------------------------------------------
Ran 2 tests in 0.000s
OK
Two successful tests isn’t surprising, since our functions don’t actually
test anything yet. The next step is to fill them in so that they actually do
something useful. The goal of testing is to confirm that our code works
properly; for to_celsius, this means that given a value in Fahrenheit, the
function produces the corresponding value in Celsius.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 73
It’s clearly not practical to try every possible value—after all, there are a
lot of real numbers. Instead, we select a few representative values and
make sure the function does the right thing for them.
For example, let’s make sure that the round-off version of to_celsius from
Section 4.2, Providing Help, on page 59 returns the right result for two
reference values: 32 Fahrenheit (0 Celsius) and 212 Fahrenheit (100
Celsius). Just to be on the safe side, we should also check a value that
doesn’t translate so neatly. For example, 100 Fahrenheit is 37.777...
Celsius, so our function should return 38 (since it’s rounding off).
We can execute each test by comparing the actual value returned by
the function with the expected value that it’s supposed to return. In
this case, we use an assert statement to let Nose know that to_celsius(100)
should be 38:
Download modules/assert.py
import nose
from temp_with_doc import to_celsius
def test_freezing():
'''Test freezing point.'''
assert to_celsius(32) == 0
def test_boiling():
'''Test boiling point.'''
assert to_celsius(212) == 100
def test_roundoff():
'''Test that roundoff works.'''
assert to_celsius(100) == 38 # NOT 37.777...
if __name__ == '__main__':
nose.runmodule()
When the code is executed, each test will have one of three outcomes:
• Pass. The actual value matches the expected value.
• Fail. The actual value is different from the expected value.
• Error. Something went wrong inside the test itself; in other words,
the test code contains a bug. In this case, the test doesn’t tell us
anything about the system being tested.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 74
Run the test module; the output should be as follows:
Download modules/outcome.out
...
----------------------------------------------------------------------
Ran 3 tests in 0.002s
OK
As before, the dots tell us that the tests are passing.
Just to prove that Nose is doing the right thing, let’s compare to_celsius’s
result with 37.8 instead:
Download modules/assert2.py
import nose
from temp_with_doc import to_celsius
def test_to_celsius():
'''Test function for to_celsius'''
assert to_celsius(100) == 37.8
if __name__ == '__main__':
nose.runmodule()
This causes the test case to fail, so the dot corresponding to it is
replaced by an “F,” an error message is printed, and the number of
failures is listed in place of OK:
Download modules/fail.out
F
======================================================================
FAIL: Test function for to_celsius
----------------------------------------------------------------------
Traceback (most recent call last):
File "/python25/lib/site-packages/nose/case.py", line 202, in runTest
self.test(*self.arg)
File "assert2.py", line 6, in test_to_celsius
assert to_celsius(100) == 37.8
AssertionError
----------------------------------------------------------------------
Ran 1 test in 0.000s
FAILED (failures=1)
The error message tells us that the failure happened in test_to_celsius
on line 6. That is helpful, but the reason for failure can be made even
clearer by adding a description of what is being tested to each assert
statement.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 75
Download modules/assert3.py
import nose
from temp_with_doc import to_celsius
def test_to_celsius():
'''Test function for to_celsius'''
assert to_celsius(100) == 37.8, 'Returning an unrounded result'
if __name__ == '__main__':
nose.runmodule()
That message is then included in the output:
Download modules/fail_comment.out
F
======================================================================
FAIL: Test function for to_celsius
----------------------------------------------------------------------
Traceback (most recent call last):
File "c:\Python25\Lib\site-packages\nose\case.py", line 202, in runTest
self.test(*self.arg)
File "assert3.py", line 6, in test_to_celsius
assert to_celsius(100) == 37.8, 'Returning an unrounded result'
AssertionError: Returning an unrounded result
----------------------------------------------------------------------
Ran 1 test in 0.000s
FAILED (failures=1)
Having tested test_to_celsius with one value, we need to decide whether
any other test cases are needed. The description of that test case states
that it is a positive value, which implies that we may also want to test
our code with a value of 0 or a negative value. The real question is
whether our code will behave differently for those values. Since all we’re
doing is some simple arithmetic, we probably don’t need to bother; in
future chapters, though, we will see functions that are complicated
enough to need several tests each.
Let’s move on to test_above_freezing. The function it is supposed to test,
above_freezing, is supposed to return True for any temperature above
freezing, so let’s make sure it does the right thing for 89.4. We should
also check that it does the right thing for a temperature below freezing,
so we’ll add a check for -42.
Finally, we should also test that the function does the right thing for the
dividing case, when the temperature is exactly freezing. Values like this
are often called boundary cases, since they lie on the boundary between
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE NOTES 76
two different possible behaviors of the function. Experience shows that
boundary cases are much more likely to contain bugs than other cases,
so it’s always worth figuring out what they are and testing them.
The test module, including comments, is now complete:
Download modules/test_freezing.py
import nose
from temp_with_doc import above_freezing
def test_above_freezing():
'''Test function for above_freezing.'''
assert above_freezing(89.4), 'A temperature above freezing.'
assert not above_freezing(-42), 'A temperature below freezing.'
assert not above_freezing(0), 'A temperature at freezing.'
if __name__ == '__main__':
nose.runmodule()
When we run it, its output is as follows:
Download modules/test_freezing.out
.
----------------------------------------------------------------------
Ran 1 test in 0.000s
OK
Whoops—Nose believes that only one test was run, even though there
are three assert statements in the file. The reason is that as far as Nose
is concerned, each function is one test. If some of those functions want
to check several things, that’s their business. The problem with this is
that as soon as one assertion fails, Python stops executing the func-
tion it’s in. As a result, if the first check in test_above_freezing failed,
we wouldn’t get any information from the ones after it. It is therefore
generally a good idea to write lots of small test functions, each of which
only checks a small number of things, rather than putting dozens of
assertions in each function.
4.6 Style Notes
Anything that can go in a Python program can go in a module, but that
doesn’t mean that anything should. If you have functions and variables
that logically belong together, you should put them in the same module.
If there isn’t some logical connection—for example, if one of the func-
tions calculates how much carbon monoxide different kinds of cars
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 77
produce, while another figures out how strong bones are given their
diameter and density—then you shouldn’t put them in one module just
because you happen to be the author of both.
Of course, people often have different opinions about what is logical
and what isn’t. Take Python’s math module, for example; should func-
tions to multiply matrices go in there too or in a separate linear algebra
module? What about basic statistical functions? Going back to the pre-
vious paragraph, should a function that calculates gas mileage go in
the same module as one that calculates carbon monoxide emissions?
You can always find a reason why two functions should not be in the
same module, but 1,000 modules with one function each are going to
be hard for people (including you) to find their way around.
As a rule of thumb, if a module has less than half a dozen things in
it, it’s probably too small, and if you can’t sum up the contents and
purpose of a module in a one- or two-sentence docstring, it’s probably
too large. These are just guidelines, though; in the end, you will have
to decide based on how more experienced programmers have organized
modules like the ones in the Python standard library and eventually on
your own sense of style.
4.7 Summary
In this chapter, we learned the following:
• A module is a collection of functions and variables grouped
together in a file. To use a module, you must first import it. After
it has been imported, you refer to its contents using modulename.
thingname.
• Put docstrings at the start of modules or functions to describe
their contents and use.
• Every “thing” in a Python program is an object. Objects have meth-
ods, which work just like functions but are associated with the
object’s type. Methods are called using object.methodname, just
like the functions in a module.
• You can manipulate images using the picture module, which has
functions for loading, displaying, and manipulating entire images,
as well as inspecting and modifying individual pixels and colors.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 78
• Programs have to do more than just run to be useful; they have
to run correctly. One way to ensure that they do is to test them,
which you can do in Python using the Nose module. Since you
usually can’t test every possible case, you should focus your test-
ing on boundary cases.
4.8 Exercises
Here are some exercises for you to try on your own:
1. Import module math, and use its functions to complete the follow-
ing exercises:
a) Write a single expression that rounds the value of -4.3 and
then takes the absolute value of that result.
b) Write an expression that takes the ceiling of sine of 34.5.
2. In the following exercises, you will work with Python’s calendar
module:
a) Visit the Python documentation website at http://docs.python.
org/modindex.html, and look at the documentation on the cal-
endar module.
b) Import the calendar module.
c) Read the description of the function isLeap. Use isLeap to de-
termine the next leap year.
d) Find and use a function in module calendar to determine how
many leap years there will be between the years 2000 and
2050, inclusive.
e) Find and use a function in module calendar to determine
which day of the week July 29, 2016 will be.
3. Using string methods, write expressions that do the following:
a) Capitalize ’boolean’.
b) Find the first occurrence of ’2’ in ’C02 H20’.
c) Find the second occurrence of "2" in ’C02 H20’.
d) Determine whether ’Boolean’ begins with a lowercase.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 79
e) Convert "MoNDaY" to lowercase letters and then capitalize the
result.
f) Remove the leading whitespace from " Monday".
4. The example used to explain import * was as follows:
Download modules/from2.cmd
>>> from math import *>>> '%6f' % sqrt(8)
'2.828427'
Explain why there are quotes around the value 2.828427.
5. Why do you think the media module mentioned in Section 4.3,
Objects and Methods, on page 60 isn’t part of the standard Python
library? How do you think Python’s developers decide what should
be in the standard library and what shouldn’t? If you need some-
thing that isn’t in the standard library, where and how can you
find it?
6. Write a program that allows the user to choose a file and then
shows the picture twice.
7. Write a program that allows the user to choose a file, sets the red
value of each pixel in the picture to 0, and shows the picture.
8. Write a program that allows the user to pick a file, halves the green
value of each pixel in the picture, and shows the picture.
9. Write a program that allows the user to pick a file and makes it
grayscale; it should calculate the average of red, green, and blue
values of each pixel and then set the red, green, and blue values
to that average.
10. Write a program that allows the user to pick a file, doubles the red
value of each pixel in the picture, and shows the picture. What
happens when a value larger than 255 is calculated?
11. Media outlets such as newspapers and TV stations sometimes
“enhance” photographs by recoloring them or digitally combine
pictures of two people to make them appear together. Do you think
they should be allowed to use only unmodified images? Given that
almost all pictures and TV footage are now digital and have to
be processed somehow for display, what would that rule actually
mean in practice?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 80
12. Suppose we want to test a function that calculates the distance
between two XY points:
Download modules/distance.py
import math
def distance(x0, y0, x1, y1):
'''Calculate the distance between (x0, y0) and (x1, y1).'''
return math.sqrt((x1 - x0) ** 2 + (y1 - y0) ** 2)
a) Unlike the rounding-off version of to_celsius, this returns a
floating-point number. Explain why this makes testing more
difficult.
b) A friend of yours suggests testing the function like this:
Download modules/test_distance.py
import nose
from distance import distance
def close(left, right):
'''Test if two floating-point values are close enough.'''
return abs(left - right) < 1.0e-6
def test_distance():
'''Test whether the distance function works correctly.'''
assert close(distance(1.0, 0.0, 1.0, 0.0), 0.0), 'Identical points fail.'
assert close(distance(0.0, 0.0, 1.0, 0.0), 1.0), 'Unit distance fails.'
if __name__ == '__main__':
nose.runmodule()
Explain what your friend is trying to do. As gently as you can,
point out two flaws in his approach.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 5
ListsUp to this point, each variable we have created has referred to a single
number or string. In this chapter, we will work with collections of data
and use a Python type named list. Lists contain 0 or more objects, and
they allow us to store data such as 90 experiment measurements or
10,000 student IDs. We’ll also see how to access files and represent
their contents using lists.
5.1 Lists and Indices
Figure 5.1, on the next page, taken from http://www.acschannelislands.
org/2008CountDaily.pdf, shows the number of gray whales counted near
the Coal Oil Point Natural Reserve in a two-week period in the spring of
2008.
Using what we have seen so far, we would have to create fourteen vari-
ables to keep track of these numbers (see Figure 5.2, on the following
page). If we wanted to track an entire year’s worth of observations, we’d
need 366 (just in case it was a leap year). Even worse, if we didn’t know
in advance how long we wanted to watch the whales, we wouldn’t know
how many variables to create.
The solution is to store all the values together in a list. Lists show up
everywhere in the real world: students in a class, the kinds of birds
native to New Guinea, and so on. To create a list in Python, we put the
values, separated by commas, inside square brackets:
Download lists/whalelist.py
# Number of whales seen per day
[5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
Prepared exclusively for Trieu Nguyen
LISTS AND INDICES 82
Day Number of Whales
1 5
2 4
3 7
4 3
5 2
6 3
7 2
8 6
9 4
10 2
11 1
12 7
13 1
14 3
Figure 5.1: Gray whale census
day1 5
day2 4
day3 7
day4 3
day5 2
day6 3
day7 2
day8 6
day9 4
day10 2
day11 1
day12 7
day13 1
day14 3
Figure 5.2: Life without lists
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LISTS AND INDICES 83
0 1 2 3
5 4 7 3
whales
4 5 6 7
2 3 2 6
8 9 10 11
4 2 1 7
12 13
1 3
Figure 5.3: List example
A list is an object; like any other object, it can be assigned to a variable:
Download lists/whales1.cmd
>>> whales = [5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
>>> whales
[5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
In Figure 5.3, we can see a memory model of whales after this assign-
ment. It’s important to keep in mind that the list itself is one object but
may contain references to other objects (shown by the arrows).
So, how do we get at the objects in a list? By providing an index that
specifies the one we want. The first item in a list is at index 0, the
second at index 1, and so on.1 To refer to a particular item, we put the
index in square brackets after a reference to the list (such as the name
of a variable):
Download lists/whales2.cmd
>>> whales = [5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
>>> whales[0]
5
>>> whales[1]
4
>>> whales[12]
1
>>> whales[13]
3
We can use only those indices that are in the range from zero up to one
less than the length of the list. In a fourteen-item list, the legal indices
are 0, 1, 2, and so on, up to 13. Trying to use an out-of-range index is
an error, just like trying to divide by zero.
1. Yes, it would be more natural to use 1 as the first index, as human languages do.
Python, however, uses the same convention as languages like C and Java and starts
counting at zero.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LISTS AND INDICES 84
Download lists/whales3.cmd
>>> whales = [5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
>>> whales[1001]
Traceback (most recent call last):
File "<stdin>", line 1, in ?
IndexError: list index out of range
Unlike most programming languages, Python also lets us index back-
ward from the end of a list. The last item is at index -1, the one before
it at index -2, and so on:
Download lists/whales4.cmd
>>> whales = [5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
>>> whales[-1]
3
>>> whales[-2]
1
>>> whales[-14]
5
We can assign the values in a list to other variables:
Download lists/whales5.cmd
>>> whales = [5, 4, 7, 3, 2, 3, 2, 6, 4, 2, 1, 7, 1, 3]
>>> third = whales[2]
>>> print 'Third day:', third
Third day: 7
The Empty List
Zero is a useful number, and as we saw in Chapter 3, Strings, on
page 39, the empty string is often useful as well. There is also an empty
list, in other words, a list with no items in it. As you might guess, it is
written [ ]. Trying to index an empty list always results in an error:
Download lists/whales6.cmd
>>> whales = []
>>> whales[0]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
IndexError: list index out of range
>>> whales[-1]
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
IndexError: list index out of range
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODIFYING LISTS 85
This follows from the definition of legal index:
• Legal indices for a list of N items are the integers in the set {i: 0 ≤
i < N }.
• The length of the empty list is 0.
• Legal indices for the empty list are therefore the elements of the
set {i: 0 ≤ i < -1}.
• Since this set is empty, there are no legal indices for the empty
list.
Lists Are Heterogeneous
Lists can contain any type of data, including integers, strings, and even
other lists. Here is a list of information about the element Krypton,
including its name, symbol, melting point (in degrees Celsius), and
boiling point (also in degrees Celsius). Using a list to aggregate related
information is somewhat prone to error; a better, but more advanced,
way to do this is described in Chapter 13, Object-Oriented Programming,
on page 270.
Download lists/krypton1.cmd
>>> krypton = ['Krypton', 'Kr', -157.2, -153.4]
>>> krypton[1]
'Kr'
>>> krypton[2]
-157.19999999999999
5.2 Modifying Lists
Suppose we’re typing in a list of the noble gases2 and our fingers slip:
Download lists/nobles1.cmd
>>> nobles = ['helium', 'none', 'argon', 'krypton', 'xenon', 'radon']
The error here is that we typed ’none’ instead of ’neon’. Rather than
retyping the whole list, we can assign a new value to a specific element
of the list:
Download lists/nobles2.cmd
>>> nobles = ['helium', 'none', 'argon', 'krypton', 'xenon', 'radon']
>>> nobles[1] = 'neon'
>>> nobles
['helium', 'neon', 'argon', 'krypton', 'xenon', 'radon']
2. A noble gas is one whose outermost electron shell is completely full, which makes it
chemically inert.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BUILT -IN FUNCTIONS ON LISTS 86
0 1 2 3
'helium' 'none' 'argon' 'krypton'
nobles
4
'xenon'
Before
5
'radon'
After
0 1 2 3
'helium' 'none' 'argon' 'krypton'
nobles
4
'xenon'
5
'radon'
'neon'
Figure 5.4: List mutation
In Figure 5.4, we show what the assignment to nobles[1] did. It also
shows that lists are mutable, in other words, that their contents can be
changed after they have been created. In contrast, numbers and strings
are immutable. You cannot, for example, change a letter in a string after
you have created it. Methods that appear to, like upper, actually create
new strings:
Download lists/strings_immutable.cmd
>>> name = 'Darwin'
>>> capitalized = name.upper()
>>> print capitalized
'DARWIN'
>>> print name
'Darwin'
The expression L[i] behaves just like a simple variable (see Section 2.4,
Variables and the Assignment Statement, on page 25). If it’s on the right,
it means “Get the value of the item at location i in the list L.” If it’s on
the left, it means “Figure out where item i in the list L is located so that
we can overwrite it.”
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BUILT -IN FUNCTIONS ON LISTS 87
Function Description
len(L) Returns the number of items in list L
max(L) Returns the maximum value in list L
min(L) Returns the minimum value in list L
sum(L) Returns the sum of the values in list L
Figure 5.5: List functions
5.3 Built-in Functions on Lists
Section 2.6, Function Basics, on page 30 introduced a few of Python’s
built-in functions. Some of these, such as len, can be applied to lists as
well, as can others that we haven’t seen before (see Figure 5.5). Here
they are in action working on a list of the half-lives3 of our plutonium
isotopes:
Download lists/plu4.cmd
>>> half_lives = [87.74, 24110.0, 6537.0, 14.4, 376000.0]
>>> len(half_lives)
5
>>> max(half_lives)
376000.0
>>> min(half_lives)
14.4
>>> sum(half_lives)
406749.14000000001
We can use the results of the built-in functions in expressions; for
example, the following code demonstrates that we can check whether
an index is in range:
Download lists/plu5.cmd
>>> half_lives = [87.74, 24110.0, 6537.0, 14.4, 376000.0]
>>> i = 2
>>> 0 <= i < len(half_lives)
True
>>> half_lives[i]
6537.0
>>> i = 5
>>> 0 <= i < len(half_lives)
False
3. The half-life of a radioactive substance is the time taken for half of it to decay. After
twice this time has gone by, three quarters of the material will have decayed; after three
times, seven eighths, and so on.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BUILT -IN FUNCTIONS ON LISTS 88
0 1 2 3
0 1
'H' 'He'
original
final
2
'Li'
0
'Be'
Figure 5.6: List concatenation
>>> half_lives[i]
Traceback (most recent call last):
File "<stdin>", line 1, in ?
IndexError: list index out of range
Like all other objects, lists have a particular type, and Python complains
if you try to combine types in inappropriate ways. Here’s what happens
if you try to “add” a list and a string:
Download lists/add_list_str.cmd
>>> ['H', 'He', 'Li'] + 'Be'
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: can only concatenate list (not "str") to list
That error report is interesting. It hints that we might be able to con-
catenate lists with lists to create new lists, just as we concatenated
strings to create new strings. A little experimentation shows that this
does in fact work:
Download lists/concat_lists.cmd
>>> original = ['H', 'He', 'Li']
>>> final = original + ['Be']
>>> final
['H', 'He', 'Li', 'Be']
As shown in Figure 5.6, this doesn’t modify either of the original lists.
Instead, it creates a new list whose entries refer to the entries of the
original lists.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PROCESSING LIST ITEMS 89
So if + works on lists, will sum work on lists of strings? After all, if sum([1,
2, 3]) is the same as 1 + 2 + 3, shouldn’t sum(’a’, ’b’, ’c’) be the same as
’a’ + ’b’ + ’c’, or ’abc’? The following code shows that the analogy can’t
be pushed that far:
Download lists/sum_of_str.cmd
>>> sum(['a', 'b', 'c'])
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: unsupported operand type(s) for +: 'int' and 'str'
On the other hand, you can multiply a list by an integer to get a new
list containing the elements from the original list repeated a certain
number of times:
Download lists/mult_lists.cmd
>>> metals = 'Fe Ni'.split()
>>> metals * 3
['Fe', 'Ni', 'Fe', 'Ni', 'Fe', 'Ni']
As with concatenation, the original list isn’t modified; instead, a new
list is created. Notice, by the way, how we use string.split to turn the
string ’Fe Ni’ into a two-element list [’Fe’, ’Ni’]. This is a common trick in
Python programs.
5.4 Processing List Items
Lists were invented so that we wouldn’t have to create 1,000 variables
to store a thousand values. For the same reason, Python has a for loop
that lets us process each element in a list in turn, without having to
write one statement per element. The general form of a for loop is as
follows:
for variable in list:
block
As we saw in Section 2.6, Function Basics, on page 30, a block is just a
sequence of one or more statements. variable and list are just a variable
and a list.
When Python encounters a loop, it executes the loop’s block once for
each value in the list. Each pass through the block is called an iteration,
and at the start of each iteration, Python assigns the next value in the
list to the specified variable. In this way, the program can do something
with each value in turn.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PROCESSING LIST ITEMS 90
For example, this code prints every velocity of a falling object in metric
and imperial units:
Download lists/velocity_loop.cmd
>>> velocities = [0.0, 9.81, 19.62, 29.43]
>>> for v in velocities:
... print "Metric:", v, "m/sec;",
... print "Imperial:", v * 3.28, "ft/sec"
...
Metric: 0.0 m/sec; Imperial: 0.0 ft/sec
Metric: 9.81 m/sec; Imperial: 32.1768 ft/sec
Metric: 19.62 m/sec; Imperial: 64.3536 ft/sec
Metric: 29.43 m/sec; Imperial: 96.5304 ft/sec
Here are two other things to notice about this loop:
• In English we would say “for each velocity in the list, print the
metric value, and then print the imperial value.” In Python, we
said roughly the same thing.
• As with function definitions, the statements in the loop block are
indented. (We use four spaces in this book; check with your in-
structors to find out whether they prefer something else.)
In this case, we created a new variable v to store the current value
taken from the list inside the loop. We could equally well have used an
existing variable. If we do this, the loop still starts with the first element
of the list—whatever value the variable had before the loop is lost:
Download lists/velocity_recycle.cmd
>>> speed = 2
>>> velocities = [0.0, 9.81, 19.62, 29.43]
>>> for speed in velocities:
... print "Metric:", speed, "m/sec;",
...
Metric: 0.0 m/sec
Metric: 9.81 m/sec
Metric: 19.62 m/sec
Metric: 29.43 m/sec
>>> print "Final:", speed
Final: 29.43
Either way, the variable is left holding its last value when the loop fin-
ishes. Notice, by the way, that the last print statement in this program
is not indented, so it is not part of the for loop. It is executed after the
for loop has finished and is executed only once.
Nested Loops
We said earlier that the block of statements inside a loop could contain
anything. This means that it can also contain another loop.Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PROCESSING LIST ITEMS 91
This program, for example, loops over the list inner once for each ele-
ment of the list outer:
Download lists/nested_loops.cmd
>>> outer = ['Li', 'Na', 'K']
>>> inner = ['F', 'Cl', 'Br']
>>> for metal in outer:
... for gas in inner:
... print metal + gas
...
...
LiF
LiCl
LiBr
NaF
NaCl
NaBr
KF
KCl
KBr
If the outer loop has No iterations and the inner loop executes Ni times
for each of them, the inner loop will execute a total of NoNi times. One
special case of this is when the inner and outer loops are running over
the same list of length N, in which case the inner loop executes N2 times.
This can be used to generate a multiplication table; after printing the
header row, we use a nested loop to print each row of the table in turn,
using tabs to make the columns line up:
Download lists/multiplication_table.py
def print_table():
'''Print the multiplication table for numbers 1 through 5.'''
numbers = [1, 2, 3, 4, 5]
# Print the header row.
for i in numbers:
print '\t' + str(i),
print # End the header row.
# Print the column number and the contents of the table.
for i in numbers:
print i,
for j in numbers:
print '\t' + str(i * j),
print # End the current row.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SLICING 92
Here is print_table’s output:
Download lists/multiplication_out.txt
>>> from multiplication_table import *>>> print_table()
1 2 3 4 5
1 1 2 3 4 5
2 2 4 6 8 10
3 3 6 9 12 15
4 4 8 12 16 20
5 5 10 15 20 25
Notice when the two different kinds of formatting are done: the print
statement at the bottom of the program prints a new line when outer
loop advances, while the inner loop includes a tab in front of each item.
5.5 Slicing
Geneticists describe C. elegans (nematodes, or microscopic worms)
using three-letter short-form markers. Examples include Emb (embry-
onic lethality), Him (High incidence of males), Unc (Uncoordinated), Dpy
(dumpy: short and fat), Sma (small), and Lon (long). We can thus keep
a list:
Download lists/celegans.cmd
>>> celegans_markers = ['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
>>> celegans_markers
['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
It turns out that Dpy worms and Sma worms are difficult to distin-
guish from each other, so they are not as useful as markers in complex
strains. We can produce a new list based on celegans_markers, but with-
out Dpy or Sma, by taking a slice of the list:
Download lists/celegans1.cmd
>>> celegans_markers = ['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
>>> useful_markers = celegans_markers[0:4]
This creates a new list consisting of only the four distinguishable mark-
ers (see Figure 5.7, on the following page).
The first index in the slice is the starting point. The second index is one
more than the index of the last item we want to include. More rigorously,
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SLICING 93
0 1 2 3
'Emb' 'Him' 'Unc' 'Lon'
celegans_markers
'Dpy'
4
1
useful_markers
0
'Sma'
5
32
Figure 5.7: Slicing doesn’t modify lists.
list[i:j] is a slice of the original list from index i (inclusive) up to, but not
including, index j (exclusive).4
The first index can be omitted if we want to slice from the beginning of
the list, and the last index can be omitted if we want to slice to the end:
Download lists/celegans2.cmd
>>> celegans_markers = ['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
>>> celegans_markers[:4]
['Emb', 'Him', 'Unc', 'Lon']
>>> celegans_markers[4:]
['Dpy', 'Sma']
To create a copy of the entire list, we just omit both indices so that the
“slice” runs from the start of the list to its end:
Download lists/celegans3.cmd
>>> celegans_markers = ['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
>>> celegans_copy = celegans_markers[:]
>>> celegans_markers[5] = 'Lvl'
>>> celegans_markers
['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Lvl']
>>> celegans_copy
['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
4. Python uses this convention to be consistent with the rule that the legal indices for a
list go from 0 up to one less than the list’s length.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ALIASING 94
0 1 2 3
'Emb' 'Him' 'Unc' 'Lon'
celegans_markers
'Dpy'
4
'Sma'
5
celegans_copy
Figure 5.8: Aliasing lists
5.6 Aliasing
An alias is an alternative name for something. In Python, two variables
are said to be aliases when they refer to the same value. For example,
the following code creates two variables, both of which refer to a single
list (see Figure 5.8). When we modify the list using one of the variables,
references through the other variable show the change as well:
Download lists/celegans4.cmd
>>> celegans_markers = ['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Sma']
>>> celegans_copy = celegans_markers
>>> celegans_markers[5] = 'Lvl'
>>> celegans_markers
['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Lvl']
>>> celegans_copy
['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Lvl']
Aliasing is one of the reasons why the notion of mutability is important.
For example, if x and y refer to the same list, then any changes you
make to the list through x will be “seen” by y, and vice versa. This can
lead to all sorts of hard-to-find errors in which a list’s value changes
as if by magic, even though your program doesn’t appear to assign
anything to it. This can’t happen with immutable values like strings.
Since a string can’t be changed after it has been created, it’s safe to
have aliases for it.
Aliasing in Function Calls
Aliasing occurs when we use list parameters as well, since parameters
are variables.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LIST METHODS 95
Method Description
L.append(v) Appends value v to list L
L.insert(i, v) Inserts value v at index i in list L, shifting following items
to make room
L.remove(v) Removes the first occurrence of value v from list L
L.reverse() Reverses the order of the values in list L
L.sort() Sorts the values in list L in ascending order (for strings,
alphabetical order)
L.pop() Removes and returns the last element of L (which must be
nonempty)
Figure 5.9: List methods
Here is a simple function that takes a list, sorts it, and then reverses it:
Download lists/alias_parameters.cmd
>>> def sort_and_reverse(L):
... '''Return list L sorted and reversed.'''
... L.sort()
... L.reverse()
... return L
...
>>> celegans_markers = ['Emb', 'Him', 'Unc', 'Lon', 'Dpy', 'Lvl']
>>> sort_and_reverse(celegans_markers)
['Unc', 'Lvl', 'Lon', 'Him', 'Emb', 'Dpy']
>>> celegans_markers
['Unc', 'Lvl', 'Lon', 'Him', 'Emb', 'Dpy']
This function modifies list L, and since L is an alias of celegans_markers,
that list is modified as well.
5.7 List Methods
Lists are objects and thus have methods. Some of the most commonly
used are listed in Figure 5.9. Here is a sample interaction showing how
we can use these methods to construct a list containing all the colors
of the rainbow:
Download lists/colors.cmd
>>> colors = 'red orange green black blue'.split()
>>> colors.append('purple')
>>> colors
['red', 'orange', 'green', 'black', 'blue', 'purple']
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LIST METHODS 96
Where Did My List Go?
Beginning programmers often forget that many list methodsreturn None rather than creating and returning a new list. (Expe-rienced programmers sometimes forget too.) As a result, theirlists sometimes seem to disappear:
Download lists/colors2.cmd
>>> colors = 'red orange yellow green blue purple'.split()>>> colors['blue', 'green', 'orange', 'purple', 'red', 'yellow']>>> sorted_colors = colors.sort()>>> print sorted_colorsNone
As we’ll discuss in Section 4.5, Testing, on page 70, mistakes likethese can quickly be caught by writing and running a few tests.
>>> colors.insert(2, 'yellow')
>>> colors
['red', 'orange', 'yellow', 'green', 'black', 'blue', 'black', 'purple']
>>> colors.remove('black')
>>> colors
['red', 'orange', 'yellow', 'green', 'blue', 'purple']
It is important to note that all these methods modify the list instead
of creating a new list. They do this because lists can grow very, very
large—a million patient records, for example, or a billion measurements
of a magnetic field. Creating a new list every time someone wanted to
make a change to such a list would slow Python down so much that it
would no longer be useful; having Python guess when it should make
a copy, and when it should operate on the list in place, would make it
impossible to figure out.
It’s just as important to remember that all of these methods except
pop return the special value None, which means “There is no useful
information” or “There’s nothing here.” Python doesn’t display anything
when asked to display the value None. Printing it, on the other hand,
shows us that it’s there:
Download lists/none.cmd
>>> x = None
>>> x
>>> print x
None
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
NESTED LISTS 97
0
life
1 2
0 1 0 1 0 1
'Canada' 80
'United States' 75.5
'Mexico' 72.0
Figure 5.10: Nested lists
Finally, a call to append is not the same as using +. First, append
appends a single value, while + expects two lists as operands. Second,
append modifies the list rather than creating a new one.
5.8 Nested Lists
We said in Section 5.1, Lists Are Heterogeneous, on page 85 that lists
can contain any type of data. That means that they can contain other
lists, just as the body of a loop can contain another loop. For example,
the following nested list describes life expectancies in different coun-
tries:
Download lists/lifelist.py
[['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0]]
As shown in Figure 5.10, each element of the outer list is itself a list
of two items. We use the standard notation to access the items in the
outer list:
Download lists/life0.cmd
>>> life = [['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0]]
>>> life[0]
['Canada', 76.5]
>>> life[1]
['United States', 75.5]
>>> life[2]
['Mexico', 72.0]
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
NESTED LISTS 98
0
life
1 2
0 1 0 1 0 1
'Canada' 76.5
'United States' 75.5
'Mexico' 72.0
canada
Figure 5.11: Aliasing sublists
Since each of these items is also a list, we can immediately index it
again, just as we can chain together method calls or pass the result of
one function call as an argument to another function:
Download lists/life1.cmd
>>> life = [['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0]]
>>> life[1]
['United States', 75.5]
>>> life[1][0]
'United States'
>>> life[1][1]
75.5
We can also assign sublists to variables:
Download lists/life2.cmd
>>> life = [['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0]]
>>> canada = life[0]
>>> canada
['Canada', 76.5]
>>> canada[0]
'Canada'
>>> canada[1]
76.5
Assigning a sublist to a variable creates an alias for that sublist (see
Figure 5.11). As before, any change we make through the sublist refer-
ence will show up when we access the main list, and vice versa:
Download lists/life3.cmd
>>> life = [['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0]]
>>> canada = life[0]
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OTHER KINDS OF SEQUENCES 99
>>> canada[1] = 80.0
>>> canada
['Canada', 80.0]
>>> life
[['Canada', 80.0], ['United States', 75.5], ['Mexico', 72.0]]
5.9 Other Kinds of Sequences
Lists aren’t the only kind of sequence in Python. You’ve already met one
of the others: strings. Formally, a string is an immutable sequence of
characters. The “sequence” part of this definition means that it can be
indexed and sliced like a list to create new strings:
Download lists/string_seq.cmd
>>> rock = 'anthracite'
>>> rock[9]
'e'
>>> rock[0:3]
'ant'
>>> rock[-5:]
'acite'
>>> for character in rock[:5]:
... print character
...
a
n
t
h
r
Python also has an immutable sequence type called a tuple. Tuples are
written using parentheses instead of square brackets; like strings and
lists, they can be subscripted, sliced, and looped over:
Download lists/tuples1.cmd
>>> bases = ('A', 'C', 'G', 'T')
... for b in bases:
... print b
A
C
G
T
There is one small catch: although () represents the empty tuple, a tuple
with one element is not written as (x) but instead as (x,) (with a trailing
comma). This has to be done to avoid ambiguity. If the trailing comma
weren’t required, (5 + 3) could mean either 8 (under the normal rules of
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FILES AS LISTS 100
arithmetic) or the tuple containing only the value 8. This is one of the
few places where Python’s syntax leaves something to be desired....
Once a tuple is created, it cannot be changed:
Download lists/life4.cmd
>>> life = (['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0])
>>> life[0] = life[1]
Traceback (most recent call last):
File "<stdin>", line 1, in ?
TypeError: object does not support item assignment
However, the objects inside it can still be changed:
Download lists/life5.cmd
>>> life = (['Canada', 76.5], ['United States', 75.5], ['Mexico', 72.0])
>>> life[0][1] = 80.0
>>> life
(['Canada', 80.0], ['United States', 75.5], ['Mexico', 72.0])
This is because it’s actually sloppy English to say that something is
“inside” a tuple. It would be more accurate to say this: “The references
contained in a tuple cannot be changed after the tuple has been cre-
ated, though the objects referred to may themselves change.”
Newcomers to Python often ask why tuples exist. The answer is that
they make some operations more efficient and others safer. We won’t
get far enough in this book to explain the former, but we will explore
the latter in Chapter 9, Sets and Dictionaries, on page 185.
5.10 Files as Lists
Most data is stored in files, which are just ordered sequences of bytes.
Those bytes may represent characters, pixels, or postal codes; the im-
portant thing is that they’re in a particular order, which means that
lists are usually a natural way to work with them.
In order to read data from a file, we must first open it using Python’s
built-in function open:
Download lists/open_basic.cmd
>>> file = open("data.txt", "r")
The first argument to open is a string containing the name of the file.
The second argument indicates a mode. The three options are "r" for
reading, "w" for writing, and "a" for appending. (The difference between
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FILES AS LISTS 101
writing and appending is that writing a file erases anything that was
already in it, while appending adds new data to the end.)
The result of open is not the contents of the file. Instead, open returns
a file object whose methods allow the program to access the contents of
the file.
The most fundamental of these methods is read. When it is called with-
out any arguments, it reads all the data in the file and returns it as a
string of characters. If we give read a positive integer argument, it reads
only up to that many characters; this is useful when we are working
with very large files. In either case, if there’s no more data in the file,
the method returns an empty string.
Although read gives us access to the bytes in a file, we usually use
higher-level methods to do our work. If the file contains text, for exam-
ple, we will probably want to process it one line at a time. To do this,
we can use the file object’s readline method, which reads the next line of
text from the file. A line is defined as being all the characters up to and
including the next end-of-line marker (see Section 3.3, Multiline Strings,
on page 43). Like read, readline returns an empty string when there’s no
more data in the file.
The neatest thing about readline is that Python calls it for us automat-
ically when a file object is used in a for loop. Assume this data is in a
file called data.txt:
Download lists/data.txt
Mercury
Venus
Earth
Mars
This program opens that file and prints the length of each line:
Download lists/fileinputloop.cmd
>>> data = open('data.txt', 'r')
>>> for line in data:
... print len(line)
...
8
6
6
5
Take a close look at the last line of output. There are only four char-
acters in the word Mars, but our program is reporting that the line
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FILES AS LISTS 102
is five characters long. The reason for this is that each of the lines
we read from the file has an end-of-line character at the end. We can
get rid of it using string.strip, which returns a copy of a string that has
leading and trailing whitespace characters (spaces, tabs, and newlines)
stripped away:
Download lists/fileinputloop2.cmd
>>> data = open('data.txt', 'r')
>>> for line in data:
... print len(line.strip())
...
7
5
5
4
This example shows the result of applying strip to a string with leading
and trailing whitespace:
Download lists/strip_basic.cmd
>>> compound = " \n Methyl butanol \n"
>>> print compound
Methyl butanol
>>> print compound.strip()
Methyl butanol
Note that the space inside the string is unaffected: string.strip takes
whitespace only off the front and end of the string.
Using string.strip, we can now produce the correct output when reading
from our file:
Download lists/fileinputloop_strip.cmd
>>> file = open('data.txt', 'r')
>>> for line in file:
... line = line.strip()
... print len(line)
...
7
5
5
4
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COMMENTS 103
Command-Line Arguments
We said earlier that the file data.txt contains the name of planets. To
finish the example, let’s go back to reading that file but display only a
certain range of the lines. We’ll provide the start and end line numbers
when we run the program. For example, we might want to read the first
three lines one time and lines 2 to 4 another time.
We can do this using command-line arguments. When we run a pro-
gram, we can send arguments to it, much like when we call a function
or method. These values end up in a special variable of the system
module sys called argv, which is just a list of the arguments (as strings).
sys.argv[0] always contains the name of the Python program being run.
In this case, it is read_lines_range.py. The rest of the command-line argu-
ments are in sys.argv[1], sys.argv[2], and so on.
Here, then, is a program that reads all the data from a file and displays
lines with line numbers within the start and end line range:
Download lists/read_lines_range.py
''' Display the lines of data.txt from the given starting line number to the
given end line number.
Usage: read_lines_range.py start_line end_line '''
import sys
if __name__ == '__main__':
# get the start and end line numbers
start_line = int(sys.argv[1])
end_line = int(sys.argv[2])
# read the lines of the file and store them in a list
data = open('data.txt', 'r')
data_list = data.readlines()
data.close()
# display lines within start to end range
for line in data_list[start_line:end_line]:
print line.strip()
5.11 Comments
The previous line-reading program is one of the longest we have seen
to date—so long, in fact, that we have added comments as well as a
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COMMENTS 104
docstring. The docstring is primarily for people who want to use the
program; it describes what the program does but not how.
Comments, on the other hand, are written for the benefit of future
developers.5 Each comment starts with the # character and runs to the
end of the line. We can put whatever we want in comments, because
Python ignores them completely.
Here are a few rules for good commenting:
• Assume your readers know as much Python as you do (for exam-
ple, don’t explain what strings are or what an assignment state-
ment does).
• Don’t comment the obvious—the following comment is not useful:
count = count + 1 # add one to count
• Many programmers leave comments beginning with “TODO” or
“FIXME” in code to remind themselves of things that need to be
written or tidied up.
• If you needed to think hard when you wrote a piece of software,
you should write a comment so that the next person doesn’t have
to do the same thinking all over again. In particular, if you develop
a program or function by writing a simple point-form description
in English, then making the points more and more specific until
they turn into code, you should keep the original points as com-
ments. (We will discuss this style of development further in Chap-
ter 10, Algorithms, on page 203.)
• Similarly, if a bug was difficult to find or if the fix is complicated,
you should write a comment to explain it. If you don’t, the next
programmer to work on that part of the program might think that
the code is needlessly complicated and undo your hard work.
• On the other hand, if you need lots of comments to explain what a
piece of code does, you should clean up the code. For example, if
you have to keep reminding readers what each of the fifteen lists
in a function are for, you should break the function into smaller
pieces, each of which works only with a few of those lists.
5. Including future versions of ourselves, who might have forgotten the details of this
program by the time a change needs to be made or a bug needs to be fixed.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 105
And here’s one more rule:
• An out-of-date comment is worse than no comment at all, so if
you change a piece of software, read the comments carefully and
fix any that are no longer accurate.
5.12 Summary
In this chapter, we learned the following:
• Lists are used to keep track of zero or more objects. We call the
objects in a list its elements and refer to them by position using
indices ranging from zero to one less than the length of the list.
• Lists are mutable, which means that their contents can be modi-
fied. Lists can contain any type of data, including other lists.
• Slicing is used to create new lists that have the same values or a
subset of the values of the originals.
• When two variables refer to the same object, we call them aliases.
• Tuples are another kind of Python sequence. Tuples are similar to
lists, except they are immutable.
• When files are opened and read, their contents are commonly
stored in lists of strings.
5.13 Exercises
Here are some exercises for you to try on your own:
1. Assign a list that contains the atomic numbers of the six alka-
line earth metals—beryllium (4), magnesium (12), calcium (20),
strontium (38), barium (56), and radium (88)—to a variable called
alkaline_earth_metals.
2. Which index contains Radium’s atomic number? Write the answer
in two ways, one using a positive index and one using a negative
index.
3. Which function tells you how many items there are in alkaline_
earth_metals?
4. Write code that returns the highest atomic number in alkaline_
earth_metals. (Hint: use one of the list methods from Figure 5.5, on
page 87.)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 106
5. What is the difference between print ’a’ and print ’a’,?
6. Write a for loop to print all the values in list half_lives from Sec-
tion 5.5, Slicing, on page 92, one per line.
7. Write a for loop to print all the values in list half_lives from Sec-
tion 5.5, Slicing, on page 92, all on a single line.
8. Consider the following statement, which creates a list of popula-
tions of countries in eastern Asia (China, DPR Korea, Hong Kong,
Mongolia, Republic of Korea, and Taiwan), in millions: country_
populations = [1295, 23, 7, 3, 47, 21]. Write a for loop that adds up all
the values and stores them in variable total. (Hint: give total an
initial value of zero, and, inside the loop body, add the population
of the current country to total.)
9. Create a list of temperatures in degrees Celsius with the values
25.2, 16.8, 31.4, 23.9, 28, 22.5, and 19.6, and assign it to a vari-
able called temps.
10. Using one of the list methods, sort temps in ascending order.
11. Using slicing, create two new lists, cool_temps and warm_temps,
which contain the temperatures below and above 20 degrees cel-
sius, respectively.
12. Using list arithmetic, recombine cool_temps and warm_temps in
into a new list called temps_in_celsius.
13. Write a for loop to convert all the values from temps_in_celsius into
Fahrenheit, and store the converted values in a new list temps_in_
fahrenheit. The list temps_in_celsius should remain unchanged.
14. Create a nested list where each element of the outer list con-
tains the atomic number and atomic weight for an alkaline earth
metal. The values are beryllium (4 and 9.012), magnesium (12 and
24.305), calcium (20 and 40.078), strontium (38 and 87.62), bar-
ium (56 and 137.327), and radium (88 and 226). Assign the list to
a variable alkaline_earth_metals.
15. Write a for loop to print all the values in alkaline_earth_metals, with
the atomic number and atomic weight for each alkaline earth
metal on a different line.
16. Write a for loop to create a new list called number_and_weight that
contains the elements of alkaline_earth_metals in the same order but
not nested.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 107
17. Suppose the file alkaline_metals.txt contains this:
4 9.012
12 24.305
20 20.078
38 87.62
56 137.327
88 226
Write a for loop to read the contents of alkaline_metals.txt, and store
it in a nested list with each element of the list contains the atomic
number and atomic weight for an element. (Hint: use string.split.)
18. Draw a memory model showing the effect of the following state-
ments:
values = [0, 1, 2]
values[1] = values
19. The following function does not have a docstring or comments.
Write enough of both to make it easy for the next person to under-
stand what the function does, and how, and then compare your
solution with those of at least two other people. How similar are
they? Why do they differ?
def mystery_function(values):
result = []
for i in range(len(values[0])):
result.append([values[0][i]])
for j in range(1, len(values)):
result[-1].append(values[j][i])
return result
20. Section 5.2, Modifying Lists, on page 85 said that strings are im-
mutable. Why might mutable strings be useful? Why do you think
Python made them immutable?
21. What happens when you sort a list that contains a mix of numbers
and strings, such as [1, ’a’, 2, ’b’]? Is this consistent with the rules
given in Chapter 3, Strings, on page 39 and Chapter 6, Making
Choices, on the next page for how comparison operators like <
work on numbers and strings? Is this the “right” thing for Python
to do, or would some other behavior be more useful?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 6
Making ChoicesThis chapter introduces another fundamental concepts of program-
ming: making choices. We have to do this whenever we want to have our
program behave differently depending on the data it’s working with. For
example, we might want to do different things depending on whether a
solution is acidic or basic.
The statements we’ll meet in this chapter for making choices are called
control flow statements, because they control the way the computer
executes programs. We have already met one control flow statement—
the loops introduced in Section 5.4, Processing List Items, on page 89—
and we will meet others in future chapters as well. Together, they are
what give programs their “personalities.”
Before we can explore control flow statements, we must introduce a
Python type that is used to represent truth and falsehood. Unlike the
integers, floating-point numbers, and strings we have already seen,
this type has only two values and three operators, but it is extremely
powerful.
6.1 Boolean Logic
In the 1840s, the mathematician George Boole showed that the classi-
cal rules of logic could be expressed in purely mathematical form using
only the two values “true” and “false.” A century later, Claude Shan-
non (later the inventor of information theory) realized that Boole’s work
could be used to optimize the design of electromechanical telephone
switches. His work led directly to the use of Boolean logic to design
computer circuits.
Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 109
In honor of Boole’s work, most modern programming languages use a
type named after him to keep track of what’s true and what isn’t.
In Python, that type is called bool (without an “e”). Unlike int and float,
which have billions of possible values, bool has only two: True and False.
True and False are values, just as much as the numbers 0 and -43.7. It
feels a little strange at first to think of them this way, since “true” and
“false” in normal speech are adjectives that we apply to other state-
ments. As we’ll see, though, treating True and False as nouns is natural
in programs.
Boolean Operators
There are only three basic Boolean operators: and, or, and not. not has
the highest precedence, followed by and, followed by or.
not is a unary operator; in other words, it is applied to just one value,
like the negation in the expression -(3 + 2). An expression involving not
produces True if the original value is False, and it produces False if the
original value is True:
Download cond/boolean_not_examples.cmd
>>> not True
False
>>> not False
True
In the previous example, instead of not True, we could simply use False;
and instead of not False, we could use True. Rather than apply not directly
to a Boolean value, we would typically apply not to a Boolean variable
or a more complex Boolean expression. The same goes for the following
examples of Boolean operators and and or, so although we apply them
to Boolean constants in the following examples, we’ll give an example
of how they are typically used at the end of this section.
and is a binary operator; the expression left and right is True if both left
and right are True, and it’s False otherwise:
Download cond/boolean_and_examples.cmd
>>> True and True
True
>>> False and False
False
>>> True and False
False
>>> False and True
False
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 110
cold
True
windy
True
True
False
False
True
(not cold) and windy
False
False
True
not (cold and windy)
False
True
True
False False False True
Figure 6.1: Relational and equality operators
or is also a binary operator. It produces True if either operand is True,
and it produces False only if both are False:
Download cond/boolean_or_examples.cmd
>>> True or True
True
>>> False or False
False
>>> True or False
True
>>> False or True
True
This definition is called inclusive or, since it allows both possibilities
as well as either. In English, the word or is also sometimes an exclu-
sive or. For example, if someone says, “You can have pizza or tandoori
chicken,” they probably don’t mean that you can have both. Like most
programming languages, Python always interprets or as inclusive. We
will see in the exercises how to create an exclusive or.
We mentioned earlier that Boolean operators are usually applied to
Boolean expressions, rather than Boolean constants. If we want to
express “It is not cold and windy” using two variables cold and windy
that contain Boolean values, we first have to decide what the ambigu-
ous English expression means: is it not cold but at the same time
windy, or is it not both cold and windy? A truth table for each alter-
native is shown in Figure 6.1, and the following code snippet shows
what they look like translated into Python:
Download cond/boolean_expression.cmd
>>> (not cold) and windy
>>> not (cold and windy)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 111
Symbol Operation
> Greater than
< Less than
>= Greater than or equal to
<= Less than or equal to
== Equal to
!= Not equal to
Figure 6.2: Relational and Equality Operators
Relational Operators
We said earlier that True and False are values. The most common way
to produce them in programs is not to write them down directly but
rather to create them in expressions. The most common way to do that
is to do a comparison using a relational operator. For example, 3<5 is
a comparison using the relational operator < whose value is True, while
13≥77 uses ≥ and has the value False.
As shown in Figure 6.2, Python has all the operators you’re used to
using. Some of them are represented using two characters instead of
one, like <= instead of ≤.
The most important representation rule is that Python uses == for
equality instead of just =, because = is used for assignment. Beginners
often mix the two up and type x = 3 when they meant to check whether
the variable x was equal to three. This always produces a syntax error,
but if you don’t know what to look for, it can be hard to spot the reason.
All relational operators are binary operators: they compare two values
and produce True or False, as appropriate. The “greater than” > and “less
than” < operators work as expected:
Download cond/relational_1.cmd
>>> 45 > 34
True
>>> 45 > 79
False
>>> 45 < 79
True
>>> 45 < 34
False
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 112
We can compare integers to floating-point numbers with any of the rela-
tional operators. Integers are automatically converted to floating point
when we do this, just as they are when we add 14 to 23.3:
Download cond/relational_2.cmd
>>> 23.1 >= 23
True
>>> 23.1 >= 23.1
True
>>> 23.1 <= 23.1
True
>>> 23.1 <= 23
False
The same holds for “equal to” and “not equal to”:
Download cond/relational_3.cmd
>>> 67.3 == 87
False
>>> 67.3 == 67
False
>>> 67.0 == 67
True
>>> 67.0 != 67
False
>>> 67.0 != 23
True
Of course, it doesn’t make much sense to compare two numbers that
you know in advance, since you would also know the result of the com-
parison. Relational operators therefore almost always involve variables,
like this:
Download cond/relational_var.cmd
>>> def positive(x):
... return x > 0
...
>>> positive(3)
True
>>> positive(-2)
False
>>> positive(0)
False
Combining Comparisons
We have now seen three types of operators: arithmetic, Boolean, and
relational.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 113
Here are the rules for combining them:
• Arithmetic operators have higher precedence than relational oper-
ators. For example, + and / are evaluated before < or >.
• Relational operators have higher precedence than Boolean opera-
tors. For example, comparisons are evaluated before and, or, and
not.
• All relational operators have the same precedence.
These rules mean that the expression 1 + 3 > 7 is evaluated as (1 + 3) >
7, not as 1 + (3 > 7). These rules also mean that you can often skip the
parentheses in complicated expressions:
Download cond/skipping_parens.cmd
>>> x = 2
>>> y = 5
>>> z = 7
>>> x < y and y < z
True
It’s usually a good idea to put the parentheses in, though, since it helps
the eye find the subexpressions and clearly communicates the order to
anyone reading your code:
Download cond/parens_included.cmd
>>> (x < y) and (y < z)
True
It’s very common in mathematics to check whether a value lies in a
certain range, in other words, that it is between two other values. You
can do this in Python by combining the comparisons with and:
Download cond/compare_range.cmd
>>> x = 3
>>> (1 < x) and (x <= 5)
True
>>> x = 7
>>> (1 < x) and (x <= 5)
False
This comes up so often, however, that Python lets you chain the
comparisons:
Download cond/chain1.cmd
>>> x = 3
>>> 1 < x <= 5
True
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 114
Most combinations work as you would expect, but there are cases that
may startle you:
Download cond/chain2.cmd
>>> 3 < 5 != True
True
>>> 3 < 5 != False
True
It seems impossible for both of these expressions to be True. However,
the first one is equivalent to this:
(3 < 5) and (5 != True)
while the second is equivalent to this:
(3 < 5) and (5 != False)
Since 5 is not True or False, the second half of each expression is True, so
the expression as a whole is True as well.
This kind of expression is an example of something that is a bad idea
even though it is legal.1 We strongly recommend that you only chain
comparisons in ways that would seem natural to a mathematician, in
other words, that you use < and <= together, or > and >= together, and
nothing else. If you’re tempted to do something else, resist. Use simple
comparisons and combine them with and in order to keep your code
readable. It’s also a good idea to use parentheses whenever you think
the expression you are writing may not be entirely clear.
Applying Boolean Operators to Integers, Floats, and Strings
We have already seen that Python converts ints to floats in mixed expres-
sions. It also converts numbers to bools, which means that the three
Boolean operators can be applied directly to numbers. When this hap-
pens, Python treats 0 and 0.0 as False and treats all other numbers as
True:
Download cond/not.cmd
>>> not 0
True
>>> not 1
False
>>> not 5
False
1. Sort of like going on a roller coaster right after eating two extra large ice cream sun-
daes back to back on a dare.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 115
3 and 0
True True
0
0 and 3
False
0
3 and 5
True True
5
Figure 6.3: Short-circuit evaluation
>>> not 34.2
False
>>> not -87
False
Things are more complicated with and and or. When Python evaluates
an expression containing either of these operators, it always does so
from left to right. As soon as it knows enough to stop evaluating, it
stops, even if some operands haven’t been looked at yet. The result is
the last thing that was evaluated, which is not necessarily either True or
False.
This is much easier to demonstrate than explain. Here are three expres-
sions involving and:
Download cond/and.cmd
>>> 0 and 3
0
>>> 3 and 0
0
>>> 3 and 5
5
In the first expression, Python sees a 0, which is equivalent to False,
and immediately stops evaluating. It doesn’t need to look at the 3 to
know that the expression as a whole is going to be false, since and is
true only if both operands are true (see Figure 6.3).
In the second expression, though, Python has to check both operands,
since knowing that the first one (the 3) isn’t false is not enough to know
what the value of the whole expression will be. Python also checks both
operands in the third expression; as you can see, it takes the value of
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 116
the last thing it checked as the value of the expression as a whole (in
this case, 5).
With or, if the first operand is considered to be true, or evaluates to
that value immediately, without even checking the second operand. The
reason for this is that Python already knows the answer: True or X is True,
regardless of the value of X.
If the first operand is equivalent to False, though, or has to check the
second operand. Its result is then that operand’s value:
Download cond/or.cmd
>>> 1 or 0
1
>>> 0 or 1
1
>>> True or 0
True
>>> 0 or False
False
>>> False or 0
0
>>> False or 18.2
18.199999999999999
(Remember, computers can’t represent all fractions exactly: the last
value in the previous code fragment is as close as it can get to 18.2.)
We claimed that if the first operand to the or operator is true, then
or evaluates to that value immediately without evaluating the second
operand. In order to show that this is what happens, try an expression
that divides by zero:
Download cond/div_zero.cmd
>>> 1 / 0
Traceback (most recent call last):
File "<string>", line 1, in <string>
ZeroDivisionError: integer division or modulo by zero
Now use that expression as the second operand to or:
Download cond/or_lazy.cmd
>>> True or 1 / 0
True
Since the first operand is true, the second operand is not evaluated, so
the computer never actually tries to divide anything by zero.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BOOLEAN LOGIC 117
There’s Such a Thing as Being Too Clever
An expression like y = x and 1/x works, but that doesn’t mean youshould use it, any more than you should use this:
result = test and first or second
as a shorthand for the following:
if test:result = first
else:result = second
Programs are meant to be readable. If you have to puzzle overa line of code or if there’s a high likelihood that someone seeingit for the first time will misunderstand it, it’s bad code, even if itruns correctly.
It’s possible to compare strings with each other, just as you would com-
pare numbers. The characters in strings are represented by integers: a
capital A, for example, is represented by 65, while a space is 32, and a
lowercase z is 172.2 Python decides which string is greater than which
by comparing corresponding characters from left to right. If the char-
acter from one string is greater than the character from the other, the
first string is greater than the second. If all the characters are the same,
the two strings are equal; if one string runs out of characters while the
comparison is being done (in other words, is shorter than the other),
then it is less. The following code fragment shows a few comparisons in
action:
Download cond/string_compare.cmd
>>> 'A' < 'a'
True
>>> 'A' > 'z'
False
>>> 'abc' < 'abd'
True
>>> 'abc' < 'abcd'
True
2. This encoding is called ASCII, which stands for “American Standard Code for Infor-
mation Interchange.” One of its quirks is that all the uppercase letters come before all
the lowercase letters, so a capital Z is less than a small a.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 118
Like zero, the empty string is equivalent to False; all other strings are
equivalent to True:
Download cond/empty_false.cmd
>>> '' and False
''
>>> 'salmon' or True
'salmon'
Python can also convert Booleans to numbers: True becomes 1, while
False becomes 0:
Download cond/truefalse.cmd
>>> False == 0
True
>>> True == 1
True
>>> True == 2
False
>>> False < True
True
This means that you can add, subtract, multiply, and divide using
Boolean values:
Download cond/bool_math.cmd
>>> 5 + True
6
>>> 7 - False
7
But “can” isn’t the same as “should”: adding True to 5, or multiplying the
temperature by current_time<NOON, will make your code much harder to
read. In practice, programmers routinely rely on conversion to Booleans
but rarely if ever use conversions in the other direction.
6.2 if Statements
The basic form of an if statement is as follows:
if condition:
block
The condition is an expression, such as name != ” or x < y. Note that
this doesn’t have to be a Boolean expression. As we discussed in Sec-
tion 6.1, Applying Boolean Operators to Integers, Floats, and Strings, on
page 114, non-Boolean values are automatically converted to True or
False when required.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 119
In particular, 0, None, the empty string ”, and the empty list [ ] all are
considered to false, while all other values that we have encountered are
considered to be true.
If the condition is true, then the statements in the block are executed;
otherwise, they are not. As with loops and functions, the block of state-
ments must be indented to show that it belongs to the if statement. If
you don’t indent properly, Python might raise an error, or worse, might
happily execute the code that you wrote but, because some statements
were not indented properly, do something you didn’t intend. We’ll briefly
explore both problems in this chapter.
Here is a table of solution categories based on pH level:
pH Level Solution Category
0–4 Strong acid
5–6 Weak acid
7 Neutral
8–9 Weak base
10–14 Strong base
We can use an if statement to print a message only when the pH level
given by the program’s user is acidic:
Download cond/if_basictrue.cmd
>>> ph = float(raw_input())
6.0
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
...
6.0 is acidic.
(Recall from Section 3.6, User Input, on page 46 that we have to convert
user input from a string to a float before doing the comparison.)
If the condition is false, the statements in the block are not executed:
Download cond/if_basicfalse.cmd
>>> ph = float(raw_input())
8.0
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
...
>>>
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 120
If we don’t indent the block, Python lets us know:
Download cond/if_indenterror.cmd
>>> ph = float(raw_input())
6.0
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
File "<stdin>", line 2
print "%s is acidic." % (ph)
^
IndentationError: expected an indented block
Since we’re using a block, we can have multiple statements, which are
executed only if the condition is true:
Download cond/if_multilinetrue.cmd
>>> ph = float(raw_input())
6.0
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
... print "You should be careful with that!"
...
6.0 is acidic.
You should be careful with that!
When we indent the first line of the block, the Python interpreter
changes its prompt to ... until the end of the block, which is signaled by
a blank line:
Download cond/if_multiline_indent_error.cmd
>>> ph = float(raw_input())
8.0
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
...
>>> print "You should be careful with that!"
You should be careful with that!
If we don’t indent the code that’s in the block, the interpreter complains:
Download cond/if_multiline_indent_error2.cmd
>>> ph = float(raw_input())
8.0
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
... print "You should be careful with that!"
File "<stdin>", line 3
print "You should be careful with that!"
^
SyntaxError: invalid syntax
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 121
If the program is in a file, then no blank line is needed. As soon as the
indentation ends, Python assumes that the block has ended as well.
This is therefore legal:
Download cond/if_multiline_indent_error3.cmd
ph = 8.0
if ph < 7.0:
print "%s is acidic." % (ph)
print "You should be careful with that!"
In practice, this slight inconsistency is never a problem, and most peo-
ple never even notice it.
Of course, sometimes there are situations where a single decision isn’t
sufficient. If there are multiple criteria to examine, there are a couple of
ways to handle it. One way is to use multiple if statements. For example,
we might print different messages depending on whether a pH level is
acidic or basic:
Download cond/multi_if.cmd
>>> ph = float(raw_input())
8.5
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
...
>>> if ph > 7.0:
... print "%s is basic." % (ph)
...
8.5 is basic.
>>>
In Figure 6.4, on the following page, we see that both conditions are
always evaluated, even though we know that only one of the blocks
can be executed. We can merge both cases by adding another condi-
tion/block pair using the elif keyword (which stands for “else if”); each
condition/block pair is called a clause:
Download cond/elif_basic.cmd
>>> ph = float(raw_input())
8.5
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
... elif ph > 7.0:
... print "%s is basic." % (ph)
...
8.5 is basic.
>>>
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 122
Trueph < 7.0
if-block #1
rest of
program
ph > 7.0True
if-block #2
False
False
Figure 6.4: if statement
The difference between the two is that the elif is checked only when
the if above it was false. In Figure 6.5, on the next page, we can see
the difference pictorially, with conditions drawn as diamonds, other
statements as rectangles, and arrows to show the flow of control.
An if statement can be followed by multiple elif clauses. This longer
example translates a chemical formula into English:
Download cond/elif_longer.cmd
>>> compound = raw_input()
CH3
>>> if compound == "H2O":
... print "Water"
... elif compound == "NH3":
... print "Ammonia"
... elif compound == "CH3":
... print "Methane"
...
Methane
>>>
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 123
Trueph < 7.0
if-block
elif-block
rest of
program
False
ph > 7.0
False
True
Figure 6.5: elif statement
If none of the conditions in a chain of if/elif statements are satisfied,
Python does not execute any of the associated blocks. This isn’t always
what we’d like, though. In our translation example, we probably want
our program to print something even if it doesn’t recognize the com-
pound. To do this, we add an else clause at the end of the chain:
Download cond/else_basic.cmd
>>> compound = raw_input()
H2SO4
>>> if compound == "H2O":
... print "Water"
... elif compound == "NH3":
... print "Ammonia"
... elif compound == "CH3":
... print "Methane"
... else:
... print "Unknown compound"
...
Unknown compound
>>>
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IF STATEMENTS 124
An if statement can have at most one else clause, and it has to be the
final clause in the statement. Notice there is no condition associated
with the else; logically, the following statement:
if condition:
if-block
else:
else-block
is the same as this:
if condition:
if-block
if not condition:
else-block
Nested if Statements
An if statement’s block can contain any type of Python statement, which
means that it can include other if statements. An if statement inside
another is called a nested if statement.
Download cond/nested_if.cmd
input = raw_input()
if len(input) > 0:
ph = float(input)
if ph < 7.0:
print "%s is acidic." % (ph)
elif ph > 7.0:
print "%s is basic." % (ph)
else:
print "%s is neutral." % (ph)
else:
print "No pH value was given!"
In this case, we ask the user to provide a pH value, which we’ll initially
receive as a string. The first, or outer, if statement checks whether the
user typed something, which determines whether we examine the value
of pH with the inner if statement.
Nested if statements are sometimes necessary, but they can get com-
plicated and difficult to understand. To describe when a statement is
executed, we have to mentally combine conditions; for example, print
"That’s acidic!" is executed only if the length of the string input is greater
than 0 and pH < 7.0 also evaluates to true.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STORING CONDITIONALS 125
6.3 Storing Conditionals
Take a look at the following line of code and guess what value is stored
in x:
Download cond/assign_bool.cmd
>>> x = 15 > 5
If you said “True”, you were right: 15 is greater than 5, so the comparison
produces True, and since that’s a value like any other, it can be assigned
to a variable.
The most common situation in which you would want to do this comes
up when translating decision tables into software. For example, sup-
pose you want to calculate someone’s risk of heart disease using the
following rules based on age and body mass index (BMI):
Age
<45 ≥45
BMI<22.0 Low Medium
≥22.0 Medium High
One way to implement this would be to use nested if statements:
if age < 45:
if bmi < 22.0:
risk = 'low'
else:
risk = 'medium'
else:
if bmi < 22.0:
risk = 'medium'
else:
risk = 'high'
The problem with this is that it’s hard to see that we’re testing exactly
the same condition in several places. If there were four thresholds each
for age and BMI, for example, we’d have sixteen inner conditions, and
it wouldn’t be obvious that they were all identical.
Here’s a better way to do this:
young = age < 45
slim = bmi < 22.0
if young:
if slim:
risk = 'low'
else
risk = 'medium'
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 126
else:
if slim:
risk = 'medium'
else:
risk = 'high'
We could also write this as follows:
young = age < 45
slim = bmi < 22.0
if young and slim:
risk = 'low'
elif young and not slim:
risk = 'medium'
elif not young and slim:
risk = 'medium'
elif not young and not slim:
risk = 'high'
We can even take advantage of the fact that False has the value 0 when
converted to an integer and that True has the value 1:
table = [['medium', 'high'],
['low', 'medium']]
young = age < 45
heavy = bmi >= 22.0
risk = table[young][heavy]
6.4 Summary
In this chapter, we learned the following:
• Python uses the Boolean values True and False to represent what
is true and what isn’t. Programs can combine these values using
three operators: not, and, and or.
• Boolean operators can also be applied to numeric values. 0 and
0.0 are equivalent to False; all other numeric values are equiva-
lent to True. When Boolean values are converted to numbers, False
becomes 0, and True becomes 1.
• Relational operators such as “equals” and “less than” compare val-
ues and produce a Boolean result.
• When different operators are combined in an expression, the order
of precedence from highest to lowest is arithmetic, relational, and
then Boolean.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 127
6.5 Exercises
Here are some exercises for you to try on your own:
1. For each of the following expressions, what value will the expres-
sion give? Verify your answers by typing the expressions into
Python.
a) True and not False
b) True or True and False
c) not True or not False
d) True and not 0
e) 52 < 52.3
f) 1 + 52 < 52.3
g) 4 != 4.0
2. Here is a possible definition of how and works:
If both operands are true, and’s result is its second value. If either
is false, and’s result is its first value.
Is this the rule that Python actually uses? If not, provide a counter
example.
3. You are given variables x and y.
a) Write an expression that evaluates to True if both variables are
True and that evaluates to False otherwise.
b) Write an expression that evaluates to True if x is False and eval-
uates to False otherwise.
c) Write an expression that evaluates to True if at least one of the
variables is True and evaluates to False otherwise.
4. Given variables full and empty, write an expression that evaluates
to True if at most one of the variables is True and evaluates to False
otherwise.
5. You want an automatic wildlife camera to switch on if the light
level is less than 0.01 or if the temperature is above freezing, but
not if both conditions are true.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 128
Your first attempt to write this is as follows:
if (light < 0.01) or (temperature > 0.0):
if (light < 0.01) and (temperature > 0.0):
pass
else:
camera.on()
A friend says that this is an exclusive or and that you could write
it more simply as follows:
if (light < 0.01) != (temperature > 0.0):
camera.on()
Is your friend right? If so, explain why. If not, give values for light
and temperature that will produce different results for the two frag-
ments of code.
6. In Section 2.7, Built-in Functions, on page 33, we saw the built-in
function abs. Given a variable x, write an expression that evalu-
ates to True if x and its absolute value are equal and evaluates
to False otherwise. Associate the resulting value with a variable
named result.
7. Write a function named different that has two parameters, a and b.
The function should return True if a and b refer to different values
and should return False otherwise.
8. You are given two float variables, population and land_area.
a) Write an if statement that will print the population if it is less
than 10,000,000.
b) Write an if statement that will print the population if it is
between 10,000,000 and 35,000,000.
c) Write an if statement that will print "Densely populated" if the
land density (number of people per unit of area) is greater
than 100.
d) Write an if statement that will print "Densely populated" if the
land density (number of people per unit of area) is greater
than 100 and that will print "Sparsely populated" otherwise.
9. Function to_celsius from Section 2.6, Function Basics, on page 30
converts from Fahrenheit to Celsius. Wikipedia, however, discus-
ses eight temperature scales: Kelvin, Celsius, Fahrenheit, Rank-
ine, Delisle, Newton, Rèaumur, and Rømer. Visit http://en.wikipedia.
org/wiki/Comparison_of_temperature_scales to read about them.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 129
a) Write a function convert_temperatures(t, source, target) that con-
verts temperature t from source units to target units, where
source and target are each one of "Kelvin", "Celsius", "Fahrenheit",
"Rankine", "Delisle", "Newton", "Reaumur", and "Romer".
Hint: on the Wikipedia page there are eight tables, each with
two columns and seven rows. That translates to an awful lot
of if statements—at least 8 * 7, because each of the eight
units can be converted to the seven other units. Possibly even
worse, if you decided to add another temperature scale, you
would need to add at least sixteen more if statements: eight to
convert from your new scale to each of the current ones and
eight to convert from the current ones to your new scale.
A better way is to choose one canonical scale, such as Celsius.
Your conversion function could work in two steps: convert
from the source scale to Celsius and then from Celsius to the
target scale.
b) Now, if you added a new temperature scale, how many if state-
ments would you need to add?
10. Assume we want to print a strong warning message if a pH value
is below 3.0 and otherwise simply report on the acidity. We try this
if statement:
Download cond/elif_wrongorder.cmd
>>> if ph < 7.0:
... print "%s is acidic." % (ph)
... elif ph < 3.0:
... print "%s is VERY acidic! Be careful." % (ph)
...
This prints the wrong message when a pH of 2.5 is entered. What
is the problem, and how can you fix it?
11. The following code displays a message(s) about the acidity of a
solution:
Download cond/acidity.cmd
ph = float(raw_input("Enter the ph level: "))
if pH < 7.0:
print "It's acidic!"
elif ph < 4.0:
print "It's a strong acid!"
a) What message(s) are displayed when the user enters 6.4?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 130
b) What message(s) are displayed when the user enters 3.6?
c) Make a small change to one line of the code so that both
messages are displayed when a value less than 4 is entered.
Why does the last example in Section 6.3, Storing Conditionals, on
page 125 check to see whether someone is heavy (that is, that their
weight exceeds the threshold) rather than light? If you wanted to write
the second assignment statement as light = bmi < 22.0, what change(s)
would you have to make to the lookup table?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 7
RepetitionThis chapter revisits another fundamental kind of control flow: repe-
tition. The for loops we saw in Section 5.4, Processing List Items, on
page 89 are the simplest way to do this, but many situations require
something a bit more powerful. Along with that “something,” we will
also encounter some of the tools that make experienced programmers
more productive and learn how Python keeps track of what it’s doing at
any moment in time.
7.1 Counted Loops
Conditional statements let us decide whether to do something; loops let
us do things many times. Section 5.4, Processing List Items, on page 89
introduced simple for loops, which can be used to do things with each
element of a sequence in turn. To refresh your memory, these loops
look like this:
for variable in <cf>list</cf>:
block
and are used like this:
Download loop/for_loop.cmd
>>> for c in 'alpha':
... print c
...
a
l
p
h
a
Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 132
30 10 22
0 1 2
values
v
30 10 22
0 1 2
values
v 60
for v in values:
v = 2 * v
Figure 7.1: Overwriting a loop iterator
But what if we want to change the elements of a list? For example,
suppose we want to double all of the values in a list. We can’t do this:
values[0] = 2 * values[0]
values[1] = 2 * values[1]
...
because we may not know the list’s length (and because we’d have to
rewrite our program every time the list’s length changed). This doesn’t
work either:
for v in values:
v = 2 * v
To see why, look at the memory model in Figure 7.1. Each loop iteration
assigned an element of values to the variable v. Doubling that value
inside the loop changes what v refers to but doesn’t change what’s in
the original list.
Ranges of Numbers
In order to do what we want, we need to know the index of the list
element we are working with. Getting that requires us to make a little
detour to examine a built-in function called range, which generates a
list of numbers.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 133
Here are some examples:
Download loop/range_basic.cmd
>>> range(1, 5)
[1, 2, 3, 4]
>>> range(1, 10)
[1, 2, 3, 4, 5, 6, 7, 8, 9]
>>> range(5, 10)
[5, 6, 7, 8, 9]
>>> range(10)
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
>>>
Notice that a call to range(start, stop) returns a list of integers from start
to the first integer before stop. This is (deliberately) consistent with the
way sequence indexing works: the expression seq[0:5] takes a slice of
seq up to, but not including, the value at index 5.
To save typing, a call to range with a single argument is equivalent to a
call to range(0, argument):
Download loop/range_single.cmd
>>> range(10)
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
>>> range(1)
[0]
>>> range(0)
[]
The result of range is a list like any other. As an example, the following
program calculates the sum of the integers from 1 to 100:
Download loop/for_rangebasic.cmd
>>> sum = 0
>>> for i in range(1, 101):
... sum += i
...
>>> sum
5050
Again, notice that the upper bound passed to range is one more than
the greatest integer we actually want.
By default, range generates numbers that increase by 1 successively—
this is called its step size. We can specify a different step size for range
with an optional third parameter.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 134
Here, we produce a list of leap years in the first half of this century:
Download loop/for_rangestepping.cmd
>>> range(2000, 2050, 4)
[2000, 2004, 2008, 2012, 2016, 2020, 2024, 2028, 2032, 2036, 2040, 2044, 2048]
The step size can also be negative, but when it is, the starting index
should be larger than the stopping index:
Download loop/for_rangestepping2.cmd
>>> range(2050, 2000, -4)
[2050, 2046, 2042, 2038, 2034, 2030, 2026, 2022, 2018, 2014, 2010, 2006, 2002]
Otherwise, range’s result will be empty:
Download loop/for_rangestepping_empty.cmd
>>> range(2050, 2000, 4)
[]
>>> range(2000, 2050, -4)
[]
Let’s return to our original goal of doubling the elements of a list. If the
list is called values, then len(values) is the number of elements it con-
tains, and the expression range(len(values)) produces a list containing
exactly the indices for values:
Download loop/range_len.cmd
>>> values = ['a', 'b', 'c']
>>> len(values)
3
>>> range(3)
[0, 1, 2]
>>> range(len(values))
[0, 1, 2]
If we use that list of indices in a for loop, we will iterate over the indices
for the list, rather than the values in the list itself:
Download loop/range_len_2.cmd
>>> values = ['a', 'b', 'c']
>>> for i in range(len(values)):
... print i
...
0
1
2
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 135
We can use this to get values out of the list:
Download loop/range_len_3.cmd
>>> values = ['a', 'b', 'c']
>>> for i in range(len(values)):
... print i, values[i]
...
0 a
1 b
2 c
and to overwrite list elements:
Download loop/range_len_4.cmd
>>> values = ['a', 'b', 'c']
>>> for i in range(len(values)):
... values[i] = 'X'
...
>>> values
['X', 'X', 'X']
We now have everything we need to solve our original problem:
Download loop/range_len_5.cmd
>>> values = [1, 2, 3]
>>> for i in range(len(values)):
... values[i] = 2 * values[i]
...
>>> values
[2, 4, 6]
We can tidy this up a bit using a combined operator:
Download loop/range_len_6.cmd
>>> values = [1, 2, 3]
>>> for i in range(len(values)):
... values[i] *= 2
...
>>> values
[2, 4, 6]
The enumerate Function
Looping over a list using its indices is such a common operation that
Python provides a built-in function called enumerate to help do it. Given
a sequence—a list, a tuple, or a string—enumerate returns a list of
pairs. The first element of each pair is an index, and the second is
the sequence’s value at that index.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 136
For example:
Download loop/enumerate.cmd
>>> for x in enumerate('abc'):
... print x
...
(0, 'a')
(1, 'b')
(2, 'c')
>>> for x in enumerate([10, 20, 30]):
... print x
...
(0, 10)
(1, 20)
(2, 30)
Using this gives us another way to write our double-the-values loop:
Download loop/enumerate_2.cmd
>>> values = [1, 2, 3]
>>> for pair in enumerate(values):
... i = pair[0]
... v = pair[1]
... values[i] = 2 * v
...
>>> values
[2, 4, 6]
This is easier to read when we write it like this:
Download loop/enumerate_3.cmd
>>> values = [1, 2, 3]
>>> for (i, v) in enumerate(values):
... values[i] = 2 * v
...
>>> values
[2, 4, 6]
What’s happening here is that Python actually allows multivalued as-
signment. If there are several variables on the left of an assignment
statement and an equal number of values on the right, Python matches
them up and does all the assignments at once:
Download loop/multi_assign.cmd
>>> x, y = 1, 2
>>> x
1
>>> y
2
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 137
This also works if the values on the right side are in a list, string, or
other kind of sequences—Python “explodes” the sequence on the right
and then assigns the elements to the variables on the left:
Download loop/multi_assign_explode.cmd
>>> first, second, third = [1, 2, 3]
>>> first
1
>>> second
2
>>> third
3
>>> first, second, third = 'abc'
>>> first
'a'
>>> second
'b'
>>> third
'c'
Knowing this, we can understand what happened in the for loop shown
earlier:
for (i, v) in enumerate(values):
values[i] = 2 * v
On the first iteration of the loop, enumerate(values) produced the tuple
(0, values[0]). Python saw that the loop was using two variables as its
indices, so it broke the tuple apart, assigning 0 to i and values[0] to v.
The next iteration produced 1, values[1], and so on.
Nested Loops, Revisited
As you saw in Section 5.4, Nested Loops, on page 90, loops can be
nested inside other loops. Here is an example where we use nested
loops to color every other line of an image black; the result of running
this code is shown in Figure 7.2, on the next page:
Download loop/for_nested.py
import media
lake = media.load_picture('lake.png')
width, height = media.get_width(lake), media.get_height(lake)
for y in range(0, height, 2): # Skip odd-numbered lines
for x in range(0, width):
p = media.get_pixel(lake, x, y)
media.set_color(p, media.black)
media.show(lake)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 138
Figure 7.2: lake_lines.png
The nested loops result in a full set of iterations for the inner loop, for
every iteration of the outer loop.
Looping Over Several Objects
It’s possible to use the same loop to interact with different objects. In
the following example, we’ll superimpose a baseball over a lake. We do
this by looping over the pixels in all the rows and columns of the base-
ball image and copying them one by one into the lake image. The top-left
corner of both images is located at (0, 0). We can position the baseball
away from the corner by adding offsets when we refer to indices in the
lake image; the result is shown in Figure 7.3, on the following page.
Download loop/multiloop.py
import media
baseball = media.load_picture('baseball.png')
lake = media.load_picture('lake.png')
width, height = media.get_width(baseball), media.get_height(baseball)
for y in range(0, height):
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
COUNTED LOOPS 139
Figure 7.3: baseball_lake.png
for x in range(0, width):
# Position the top-left of the baseball at (50, 25)
from_p = media.get_pixel(baseball, x, y)
to_p = media.get_pixel(lake, 50 + x, 25 + y)
media.set_color(to_p, media.get_color(from_p))
media.show(lake)
Ragged Lists
Nothing says that nested lists all have to be the same length:
Download loop/different_lengths.cmd
>>> info = [['Isaac Newton', 1643, 1727],
['Charles Darwin', 1809, 1882],
['Alan Turing', 1912, 1954, '[email protected]']]
>>> for item in info:
... print len(item)
...
3
3
4
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 140
These sorts of lists are called ragged lists. Ragged lists can be tricky to
process if the data is not uniform; for example, trying to assemble a list
of email addresses for data where some addresses are missing requires
a bit of careful thought.
Ragged data does arise normally. For example, if a record is made
each day of the time at which a person who is trying to quit smokes
a cigarette, each day will have a different number of entries. The code
is fairly straightforward:
Download loop/nested_for.cmd
>>> times = [["9:02", "10:17", "13:52", "18:23", "21:31"],
... ["8:45", "12:44", "14:52", "22:17"],
... ["8:55", "11:11", "12:34", "13:46", "15:52", "17:08", "21:15"],
... ["9:15", "11:44", "16:28"],
... ["10:01", "13:33", "16:45", "19:00"],
... ["9:34", "11:16", "15:52", "20:37"],
... ["9:01", "12:24", "18:51", "23:13"]]
>>> for day in times:
... for time in day:
... print time,
...
9:02 10:17 13:52 18:23 21:31
8:45 12:44 14:52 22:17
8:55 11:11 12:34 13:46 15:52 17:08 21:15
9:15 11:44 16:28
10:01 13:33 16:45 19:00
9:34 11:16 15:52 20:37
9:01 12:24 18:51 23:13
7.2 while Loops
for loops are very useful if you know how many iterations of the loop you
need. However, there are situations where it’s impossible to know in
advance how many times you will want the loop body executed. That’s
what a while loop is used for. while loops are sometimes called conditional
loops, since they iterate only as long as some condition is true. Their
general form is as follows:
while condition:
block
The condition of a while loop is an expression, just like the condition
of an if statement. When Python encounters a while loop, it evaluates
the condition. If that condition is false, Python skips the loop body.
If the condition is true, on the other hand, Python executes the loop
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 141
Truerabbits
> 0 block
rest of
program
False
Figure 7.4: while loops
body once and then goes back to the top of the loop and reevaluates
the condition. If it’s still true, the loop body is executed again. This is
repeated—condition, body, condition, body—until the condition is false,
at which point Python stops executing the loop.
Here’s a trivial example:
Download loop/simple_while.cmd
>>> rabbits = 3
>>> while rabbits > 0:
... print rabbits
... rabbits -= 1
...
3
2
1
Notice that this loop did not print 0. When the number of rabbits
reaches zero, the loop expression is false, so the body is not executed
(see Figure 7.4).
As a more useful example, we can calculate the growth of a bacterial
colony using a simple exponential growth model, which is essentially a
calculation of compound interest:
P(t + 1) = P(t) + rP(t)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 142
In this formula, P(t) is the population size at time t, and r is the growth
rate. Let’s see how long it takes the bacteria to double their numbers:
Download loop/whileloop.py
time = 0
population = 1000 # 1000 bacteria to start with
growth_rate = 0.21 # 21% growth per minute
while population < 2000:
population = population + growth_rate * population
print population
time = time + 1
print "It took %d minutes for the bacteria to double." % time
print "...and the final population was %6.2f bacteria." % population
Because the time variable was updated inside the loop, its value after
the loop was the time of the last iteration, which is exactly what we
want. Running this program gives us the answer we were looking for:
Download loop/whileloop_output.cmd
1210.0
1464.1
1771.561
2143.58881
It took 4 minutes for the bacteria to double.
...and the final population was 2143.59 bacteria.
Infinite Loops
The preceding example used population < 2000 as a loop condition so that
the loop stopped when the population reached double its initial size or
more. What would happen if we stopped only when the population was
exactly double its initial size?
Download loop/while_infinite.py
# Use multi-valued assignment to set up controls.
time, population, growth_rate = 0, 1000, 0.21
# Don't stop until we're exactly double original size.
while population != 2000:
population = population + growth_rate * population
print population
time = time + 1
print "It took %d minutes for the bacteria to double." % time
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 143
Here is this program’s output:
Download loop/while_infinite_output.cmd
1210.0
1464.1
1771.561
2143.58881
2593.7424601
...3,680 lines or so later...
inf
inf
inf
...and so on forever...
Whoops—since the population is never exactly 2,000 bacteria, the loop
never stops.1 A loop like this one is called an infinite loop, because
the computer will execute it forever (or until you kill your program,
whichever comes first). In Wing 101, you kill your program by selecting
Restart Shell from the Options menu, and from the command-line shell,
you can kill it by pressing Ctrl-C. Infinite loops are a common kind of
bug; the usual symptoms include printing the same value over and over
again, or hanging (doing nothing at all).
Here’s a more subtle example of an infinite loop. The following function
counts the number of times a fragment occurs in a string of DNA. It
starts searching at index 0. Subsequent searches start at the index of
the last match:
Download loop/debug.py
def count_fragments(fragment, dna):
count = -1
last_match = 0
while last_match != -1:
count += 1
last_match = dna.find(fragment, last_match)
return count
Let’s try calling it with some bits of DNA:
Download loop/debug.cmd
>>> count_fragments('atc', 'gttacgtggatg')
0
>>> count_fragments('gtg', 'gttacgtggatg')
1. The first set of dots represents more than 3,000 values, each 21 percent larger than
the one before. Eventually, these values are too large for the computer to represent, so it
displays inf (or on some computers 1.#INF), which is its way of saying “effectively infinity.”
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 144
Figure 7.5: The Wing 101 debugger
It works when the fragment doesn’t occur in the DNA. However, when
the fragment does appear, the function seems to run forever. Let’s try
to diagnose the problem using a debugger. Wing 101 has one (see Fig-
ure 7.5); so should you.
The easiest way to figure out what’s going on is to set a breakpoint
where the function is called. In Wing 101, you can do this by clicking
in the margin directly to the left of the line, as illustrated in Figure 7.6,
on the next page.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 145
Figure 7.6: Setting a breakpoint
Figure 7.7: The Stack Data tab
Start the debugging session by clicking the Debug button. You’ll notice
that the line with the breakpoint gets highlighted. Click the Step Into
button to make the debugger look at the details of the function call.
Now that we’re inside, we can step over individual statements. We don’t
want to step into any more functions at this level because we’re con-
cerned only about what’s happening in our own function.
Keep stepping over the code until you reach the beginning of the while
loop. While you’re stepping, take a look at the Stack Data tab (see Fig-
ure 7.7). You can see the values of your local variables there.
Step over the statements in the loop a few times while watching how the
values of the local variables change. Notice that count keeps increasing,
but last_match doesn’t change! It turns out that after we find the first
match, we find it over and over again in subsequent searches.
Edit the value of last_match in the Stack Data tab by double-clicking
it. Change it from 5 to 6, and continue stepping. That should let the
function run to completion.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 146
Now that we confirmed what our bug is, we need to go back and change
our code. What we really want to do is search from last_match + 1:
Download loop/debug2.py
def count_fragments(fragment, dna):
count = -1
last_match = 0
while last_match != -1:
count += 1
last_match = dna.find(fragment, last_match + 1)
return count
Now let’s try running it with different input:
Download loop/debug2.cmd
>>> count_fragments('gtg', 'gttacgtggatg')
1
>>> count_fragments('gtt', 'gttacgtggatg')
0
The first case works. The second case is obviously wrong because ’gtt’
occurs right at the beginning of the string. It turns out that we inserted
a bug along with our fix! Now that you know how to use the debugger,
you can diagnose this problem on your own.
The Call Stack
Python keeps track of any running functions using a runtime stack (see
Figure 7.8, on the next page). We can think of the stack as a series
of records, called frames (see Figure 7.9, on the following page), which
pile up on top of each other as functions are called. Only the top frame
is active; the rest are paused, waiting until functions above them are
finished.
Code outside of any function is executed in a special frame that is at
the bottom of the stack; this frame is called <module> because it is
executing at the module level.
When a function is called, Python creates a new frame for it and adds
the frame to the top of the stack. These frames store information about
each function call and the order they were called in. The most recently
called function’s frame always sits at the top of the stack.
As we will see later in this chapter, each stack frame stores a func-
tion’s parameters and local variables. It also contains a reference to the
next statement Python will execute when the function finishes. This is
known as the return address.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WHILE LOOPS 147
__main__
Runtime stack when
Python first starts
__main__
Runtime stack after
calling func1
func1
Stack grows in this directionActive frame
__main__
Runtime stack after
func1 calls func2
func2
func1
Figure 7.8: Runtime stack
argument1argument2
local1local2
__main__
FrameFunction name
Arguments
Local variables
Figure 7.9: A stack frame
Finally, the frame has space set aside for storing the function’s return
value, which we will see later in the chapter.
When a function finishes executing, Python then checks the stack to
see what it should run next. This information is stored in the return
address of the frame at the top of the stack. Once Python knows where
to go, it removes the frame.
When your Python program is finished executing, there will be no more
frames on the stack.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USER INPUT LOOPS 148
7.3 User Input Loops
We can use the raw_input function in a loop to make the chemical for-
mula translation example from Section 6.2, if Statements, on page 118
interactive:
Download loop/userinputloop_infinite.py
while True:
formula = raw_input("Please enter a chemical formula: ")
if formula == "H2O":
print "Water"
elif formula == "NH3":
print "Ammonia"
elif formula == "CH3":
print "Methane"
else:
print "Unknown compound"
In this case, we’d like the loop to continually process user input, so the
loop condition is just True. Running the program looks like this:
Download loop/userinputloop_infinite_output.cmd
Please enter a chemical formula: NH3
Ammonia
Please enter a chemical formula: H2O
Water
Please enter a chemical formula: NaCl
Unknown compound
...
where ... shows that the cycle repeats until the user kills the program
or turns off her computer. Since that isn’t a particularly friendly user
interface, let’s modify the loop condition to give users a way to stop the
program cleanly:
Download loop/userinputloop.py
text = ""
while text != "quit":
text = raw_input("Please enter a chemical formula (or 'quit' to exit): ")
if text == "quit":
print "...exiting program"
elif text == "H2O":
print "Water"
elif text == "NH3":
print "Ammonia"
elif text == "CH3":
print "Methane"
else:
print "Unknown compound"
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CONTROLLING LOOPS 149
Since the loop condition checks the value of text, we have to assign it a
value before the loop begins. Now we can run the program as usual but
exit whenever we want:
Download loop/userinputloop_output.cmd
Please enter a chemical formula (or 'quit' to exit): CH3
Methane
Please enter a chemical formula (or 'quit' to exit): H2O
Water
Please enter a chemical formula (or 'quit' to exit): quit
...exiting program
7.4 Controlling Loops
As a rule, for and while loops execute all the statements in their body
on each iteration. However, it is sometimes handy to be able to break
that rule. Python provides two ways of controlling the iteration of a loop:
break, which exits the loop body immediately, and continue, which skips
ahead to the next iteration.
The break Statement
Sometimes a loop’s task is finished before its final iteration. Using what
we have seen so far, though, we still have to finish iterating. For exam-
ple, to find which line in a file contains the string “Earth,” we would
have to write something like this:
Download loop/filesearch_nobreak.py
current_line = 1
earth_line = 0
file = open("data.txt", "r")
for line in file:
line = line.strip()
if line == "Earth":
earth_line = current_line
current_line = current_line + 1
print "Earth is at line %d" % earth_line
Here we require two variables: one for the current line number and
another to remember the desired line number until after the loop. This
is a little bit clumsy and also inefficient: if “Earth” is the first line of the
file, this program is still going to read everything that comes after it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CONTROLLING LOOPS 150
Truehas more
lines?
rest of
program
Falseline ==
'Earth'
True
False
break
Figure 7.10: The break statement
To fix this, we can exit the loop early using a break statement, which
jumps out of the loop body immediately:
Download loop/filesearch_break.py
earth_line = 1
file = open("data.txt", "r")
for line in file:
line = line.strip()
if line == "Earth":
break
earth_line = earth_line + 1
print "Earth is at line %d" % earth_line
We can see this in Figure 7.10. Notice that because the loop exits early,
the counter variable earth_line retains the value it had when the break
was executed, so we need only one variable. However, if the file con-
tains more than one occurrence of “Earth,” this program will produce
a different answer than the previous one. This program will display the
index of the first line found, while the previous program would remem-
ber and display the index of the last matching line.
One more thing about break: it exits only the innermost loop that it’s
contained in. This means that in a nested loop, a break statement inside
the inner loop will exit only the inner loop, not both loops.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CONTROLLING LOOPS 151
The continue Statement
Another way to bend the rules for iteration is to use the continue state-
ment, which causes Python to skip immediately ahead to the next iter-
ation of a loop. For example, let’s assume that the file data.txt could
include comment lines starting with a # character:
Download loop/fileinput_continue_data.cmd
# Pluto is only 0.002 times the mass of Earth.
Pluto
Mercury
# Mars is half Earth's diameter, but only
# 0.11 times Earth's mass.
Mars
Venus
Earth
Uranus
If we used the previous version of this program, it would count the
commented lines as well, giving us an incorrect answer. Instead, we
can use continue to skip comments in the file:
Download loop/filesearch_continue.cmd
entry_number = 1
file = open("data.txt", "r")
for line in file :
line = line.strip()
if line.startswith("#"):
continue
if line == "Earth":
break
entry_number = entry_number + 1
print "Earth is the %dth-lightest planet." % (entry_number)
When continue is executed, it immediately begins the next iteration of
the loop and skips any statements in the loop body that appear after it.
This allows us to skip the statement where entry_number is incremented.
Using continue is one way to skip comment lines, but this can also
be accomplished by using the conditional statements that were intro-
duced earlier in this chapter. In the previous code, continue prevents
the comment line from being processed; in other words, if the line is
not a comment, it should be processed.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CONTROLLING LOOPS 152
Truehas more
lines?
rest of
program
False
line ==
'Earth'
True
False
break
starts with
"#" ? continue
False
True
Figure 7.11: The continue statement
The form of the previous sentence matches that of an if statement, and
the updated code is as follows:
Download loop/filesearch_if.py
entry_number = 1
file = open("data.txt", "r")
for line in file :
line = line.strip()
if not line.startswith("#"):
if line == "Earth":
break
entry_number = entry_number + 1
print "Earth is the %dth-lightest planet." % (entry_number)
Whether you favor if or continue is largely a matter of personal taste.2
As always, the most important thing is to be consistent: switching back
and forth between the two is more likely to confuse your readers than
consistently using either one.
2. That is our way of saying that the authors and reviewers of this book split almost
evenly as to which they thought was better.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE NOTES 153
7.5 Style Notes
Conditional statements give us control over of the flow of execution of
our programs. When there are multiple paths of execution, we have
choices to make about the structure of our conditional statements. For
example, consider the following:
Download loop/complex_cond.py
def f(a, b, c):
if a:
if b:
print 'hi'
elif c:
print 'bonjour'
else:
print 'hola'
else:
print 'Select a language.'
Under what conditions will “hi” be printed? “hola”? The word “hi” is
printed when both a and b are True, and “hola” is printed when a is
True, but b and c are not. To make this clearer, we can rewrite our
conditional statement without the nesting:
Download loop/complex_cond_2.py
def f(a, b, c):
if a and b:
print 'hi'
elif a and c:
print 'bonjour'
elif a:
print 'hola'
else:
print 'Select a language'
With multiple ways to write equivalent conditional statements, care
must be taken to structure our code so that it is easy to understand.
Reducing the amount of nesting is one way to improve the readability
of the code, particularly if there are several nested conditionals.
break and continue have their place but should be used sparingly since
they can make programs harder to understand. When people see while
and for loops in programs, their first assumption is that the whole body
will execute every time, in other words, that the body can be treated as
a single “super statement” when trying to understand the program. If
the loop contains break or continue, though, that assumption is false.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 154
Sometimes, only part of the statement body will execute, which means
the reader has to keep two scenarios in mind.
There are always alternatives: well-chosen loop conditions (as in Sec-
tion 7.3, User Input Loops, on page 148) can replace break, and if state-
ments can be used to skip statements instead of continue. It is up to
the programmer to decide which option makes the program clearer and
which makes it more complicated. As we said in Section 5.11, Com-
ments, on page 103, programs are written for human beings; taking a
few moments to make your code as clear as possible, or to make clarity
a habit, will pay dividends for the lifetime of the program.
7.6 Summary
In this chapter, we learned the following:
• Program statements in Python can be grouped into blocks using
indentation.
• Choosing whether to execute a block is one of the fundamental
ways to control a program’s behavior. In Python, such choices are
expressed using if, elif, and else. The first two base their decision
on the value of a Boolean expression, while else is executed only if
all other tests fail.
• Repeating a block is another fundamental way to control a pro-
gram’s behavior. Using a for loop to iterate over the elements of
a structure is one example, but for can also be used to create
counted loops that iterate over a range of integers.
• The most general kind of repetition is the while loop, which contin-
ues executing as long as some arbitrary Boolean condition is true.
However, the condition is tested only at the top of the loop. If that
condition is never false, the loop will execute forever. The break and
continue statements can be used to change the way loops execute.
• Control structures like loops and conditionals can be nested inside
one another to any desired depth.
• Python and other languages keep track of nested function calls
using a call stack. Each time a function is called, a new frame
containing that functions parameters and local variables is put
on the top of the stack. When the function returns, the frame is
discarded.
• Programs can use raw_input to get input from users interactively.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 155
7.7 Exercises
Here are some exercises for you to try on your own:
1. Your lab partner claims to have written a function that replaces
each value in a list with twice the preceding value (and the first
value with 0). For example, if the list [1, 2, 3] is passed as an argu-
ment, the function is supposed to turn it into [0, 2, 4]. Here’s the
code:
Download loop/buggy_scan.py
def double_preceding(values):
if values == []:
pass # do nothing to the empty list
else:
temp = values[0]
values[0] = 0
for i in range(1, len(values)):
values[i] = 2 * temp
temp = values[i]
Explain what the bug in this function is, and fix it.
2. You are given two lists, rat_1 and rat_2, that contain the daily
weights of two rats over a period of ten days. Write statements
to do the following:
a) If the weight of Rat 1 is greater than that of Rat 2 on day 1,
print "Rat 1 weighed more than Rat 2 on Day 1."; otherwise, print
"Rat 1 weighed less than Rat 2 on Day 1."
b) If Rat 1 weighed more than Rat 2 on day 1 and if Rat 1 weighs
more than Rat 2 on the last day, print "Rat 1 remained heavier
than Rat 2."; otherwise, print "Rat 2 became heavier than Rat 1."
c) If your solution to the previous question used nested if state-
ments, then do it without nesting, or vice versa.
3. Print the numbers in the range 33 to 49 (inclusive).
4. Print the numbers from 1 to 10 in descending order, all on one
line.
5. Calculate the average of numbers in the range 2 to 22 using a loop
to find the total, and then calculate the average.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 156
6. Consider the following function:
Download loop/shrinking_list.py
def remove_neg(num_list):
'''Remove the negative numbers from the list num_list.'''
for item in num_list:
if item < 0:
num_list.remove(item)
When remove_negs([1, 2, 3, -3, 6, -1, -3, 1]) is executed, the result is [1,
2, 3, 6, -3, 1]. The for loop traverses the elements of the list, and when
a negative value (like -3 at position 3) is reached, it is removed,
shifting the subsequent values one position earlier in the list (so 6
moves into position 3). The loop then continues on to process the
next item, skipping over the value that moved into the removed
item’s position. If there are two negative numbers in a row (like -1
and -3), then the second one will not be removed. Rewrite the code
to avoid this problem.
7. Using nested for loops, print a right triangle of the character T on
the screen where the triangle is one character wide at its narrow-
est point and seven characters wide at its widest point:
T
TT
TTT
TTTT
TTTTT
TTTTTT
TTTTTTT
8. Using nested for loops, print the triangle described in the previous
question with its hypotenuse on the left side:
T
TT
TTT
TTTT
TTTTT
TTTTTT
TTTTTTT
9. Redo the previous two questions using while loops instead of for
loops.
10. The variables rat_1_weight and rat_2_weight store the weights of two
rats at the beginning of an experiment. The variables rat_1_rate
and rat_2_rate are the rate that the rats’ weights are expected to
increase each week (for example, 4 percent per week).
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 157
a) Using a while loop, calculate how many weeks it would take
for the weight of the first rat to become 25 percent heavier
than it was originally.
b) Assume that the two rats have the same initial weight, but rat
1 is expected to gain weight at a faster rate than rat 2. Using
a while loop, calculate how many weeks it would take for rat 1
to be 10 percent heavier than rat 2.
11. Transformations are some of the most common operations in
image-processing software. Arbitrary rotations require some trig-
onometry but 90 degree rotations can be done using for loops.
Design a function that uses nested for loops to rotate an image 90
degrees clockwise.
12. Reflections are another type of image transformation. Reflected
images appear flipped over when compared to the original.
a) When you look into a typical mirror, the image you see is a
horizontal reflection of reality. Use nested for loops to trans-
form an image in the same way by flipping it along the y-axis.
b) Use nested for loops to flip an image along the x-axis.
13. Scaling is another common image transformation. Downscaling
makes an image smaller, while upscaling makes it larger. Create a
function that uses nested for loops to downscale an image so that
each of its new dimensions are half of the original. For example,
your function should transform a 100 x 50 pixel image into a 50
x 25. The new image should look like it is 1/4th the size of the
original.
14. The mosaic filter is a common artistic effect in image processing
programs. The simple version of this filter divides an image into a
grid of equal-sized squares. The color of each square is determined
by the pixels that make up the square.
To keep things simple, assume each square is 10 pixels wide. The
color of the square is the average of the colors of the 100 pixels
inside. The average of two colors can be computed like this: red =
(red_1 + red_2)/2, green = (green_1 + green_2)/2, blue = (blue_1
+ blue_2)/2. Assuming the original image dimensions are divis-
ible by 10, design a function that implements the mosaic filter
described here.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 158
15. When the red, green, and blue components of a pixel are equal,
the observed color is black, white, or a shade of gray. You can
transform a color image into a grayscale image by converting each
pixel to a shade of gray that approximates the brightness of the
original color. The easiest way to do this is to replace each color
component for a pixel with the average of all the color components.
a) Design a function that uses nested for loops that transforms
a color image into grayscale using the technique described
earlier.
b) The method of producing grayscale images in the previous
exercise assumes the human eye is equally sensitive to each
of the three color components. In reality, this is not the case.
The human eye is much more sensitive to greens than the
other colors. How would you modify your solution so that
greens influence the brightness of the image more than the
reds and blues?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 8
File ProcessingWe now have the tools we need—functions, modules, lists, loops, and
conditionals—to solve problems that scientists face every day. Scientific
data is often stored in plain-text files, which can be organized in several
different ways. The most straightforward format is one piece of data per
line; for example, the rainfall in Oregon for each separate day in a study
period might be stored on a line of its own. Alternatively, each line might
store the values for an entire week or month, with a delimiter such as a
space, tab, or comma used to separate values to make the data easier
for humans to read.
Often, data is more complex. For example, a study might keep track
of the heights, weights, and ages of the participants. Each record can
appear on a line by itself with the pieces of data in each record sep-
arated by delimiters. Some records might even span multiple lines, in
which case the format will usually have some kind of a separator (such
as a blank line) between records or use special symbols to mark the
start or end of each record.
Modules to read (or parse) many common formats are part of the stan-
dard Python library or will have been written by whoever created the
format you are working with. However, it is still common to encounter
specialized formats for which readers don’t exist. This chapter will show
you how to handle a variety of file formats from fairly simple single-line
records to more complex multiline records, as well as several variations
of each.
Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 160
8.1 One Record per Line
The following data, taken from the Time Series Data Library (TSDL)
[Hynnd], describes the number of colored fox fur pelts produced in
Hopedale, Labrador, in the years 1834 –1842. (The full data set has
values for the years 1834 –1925.)
Download fileproc/hopedale.txt
Coloured fox fur production, HOPEDALE, Labrador, 1834-1842
#Source: C. Elton (1942) "Voles, Mice and Lemmings", Oxford Univ. Press
#Table 17, p.265--266
22
29
2
16
12
35
8
83
166
The first line contains a description of the data. The next two lines con-
tain comments about the data, each of which begins with a # character.
Each piece of actual data appears on a single line.
Reading a File
For now, we will treat all lines the same (data, headers, and comments)
and read all lines of the file. As we saw in Section 5.10, Files as Lists,
on page 100, this loop reads and prints a file line by line:
Download fileproc/read_file_1.py
input_file = open("hopedale.txt", "r")
for line in input_file:
line = line.strip()
print line
input_file.close()
Files Over the Internet
These days, of course, the file containing the data we want could well
be on a machine half a world away. Provided it is accessible over the
Internet, though, it is just as easy to work with. The module urllib con-
tains a function called urlopen that opens a web page for reading and
returns a file-like object that you can use exactly as if you were reading
a local file.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 161
For example, the Hopedale data not only exists on our machine, but it’s
also on a web page.1 We can print this web page line by line using a
module called urllib and commands that are very similar to those we’ve
use with files:
Download fileproc/read_url.py
import urllib
url = "http://www-personal.buseco.monash.edu.au/~hyndman/TSDL/ecology1/hopedale.dat"
web_page = urllib.urlopen(url)
for line in web_page:
line = line.strip()
print line
web_page.close()
Note that this example will work only if your machine is actually con-
nected to the Internet. Let’s return to our example of reading and print-
ing a file line by line:
Download fileproc/read_file_1.py
input_file = open("hopedale.txt", "r")
for line in input_file:
line = line.strip()
print line
input_file.close()
We can improve this in several ways. One big issue is that the filename
is hard-coded; in other words, the name of a particular file is stored
in the program, which means the program can’t be used to process
any other files. To fix this, we use what we learned in Section 5.10,
Command-Line Arguments, on page 103; we place the code that does
the work in a function that takes the name of the file as a parameter
and call that function with sys.argv[1] as an argument:
Download fileproc/read_file_2.py
import sys
def process_file(filename):
'''Open, read, and print a file.'''
input_file = open(filename, "r")
for line in input_file:
line = line.strip()
print line
1. At the time of writing, the URL for the file is
http://www-personal.buseco.monash.edu.au/~hyndman/TSDL/ecology1/hopedale.dat (you can
look at it online!).
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 162
input_file.close()
if __name__ == "__main__":
process_file(sys.argv[1])
This is more flexible, because we can now pass the program the name
of a file when we run it. To do this in Wing 101, right-click the edit-
ing pane, select Properties, select the Debug tab, and enter the run
arguments. However, it will work only with files, not with other input
streams such as standard input or a URL using module urllib. We fix
this by opening the file (or URL or input stream) outside of the function
and passing in the file object instead:
Download fileproc/read_file_3.py
import sys
def process_file(reader):
'''Read and print the contents of reader.'''
for line in reader:
line = line.strip()
print line
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
process_file(input_file)
input_file.close()
This change allows us to call the same function with an open web page
instead of a local file:
Download fileproc/read_file_4.py
import urllib
def process_file(reader):
'''Read and print the contents of reader.'''
for line in reader:
line = line.strip()
print line
if __name__ == "__main__":
webpage = urllib.urlopen(sys.argv[1])
process_file(webpage)
webpage.close()
Skipping the Header
Of course, we usually don’t just want to print data—if we did, we could
open the file in an editor and read it. Instead, we typically want to find
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 163
the smallest value, sum the values, or process it in some other way.
Before we can do that, though, we need to figure out how to skip the
file’s header. To do this, we need to read the description line and all the
comment lines following it. Of course, not all files start with a descrip-
tion line like this one does, so we won’t always want to discard the first
line. We can stop when we read the first real piece of data, which will
be the first line after the description that doesn’t start with a #:
In English, we might try this algorithm to process such a file:
Skip the first line in the file.
Skip over the comment lines in the file.
For each of the remaining lines in the file:
Process the data on that line.
The problem with this approach is that we can’t tell whether a line is
a comment line until we’ve read it, but we can read a line from a file
only once—there’s no simple way to “back up” in the file. An alternative
approach is to read the line, skip it if it’s a comment, and process it
if it’s not. Once we’ve processed the first line of data, we process the
remaining lines:
Skip the first line in the file.
Find and process the first line of data in the file.
For each of the remaining lines:
Process the data on that line.
The thing to notice about this algorithm is that it processes lines in two
places: once when it finds the first “interesting” line in the file and once
when it handles all of the following lines:
Download fileproc/tsdl.py
import sys
def skip_header(r):
'''Skip the header in reader r, and return the first
real piece of data.'''
# Read the description line and then the comment lines.
line = r.readline()
line = r.readline()
while line.startswith('#'):
line = r.readline()
# Now line contains the first real piece of data.
return line
def process_file(r):
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 164
'''Read and print open reader r.'''
# Find the first piece of data.
line = skip_header(r).strip()
print line
# Read the rest of the data.
for line in r:
line = line.strip()
print line
if __name__ == "__main__":
input_file = open(sys.argv[1], 'r')
process_file(input_file)
input_file.close()
In skip_header, we return the first line of read data, because once we’ve
found it, we can’t read it again (we can go forward but not backward).
We will want to use skip_header in all of the file-processing functions in
this section. Rather than copying the code each time we want to use
it, we can put the function in a file called tsdl.py for Time Series Data
Library and get it in other programs using import tsdl, as shown in the
next example. This allows us to reuse the skip_header code, and if it
needs to be modified, then there is only one copy of the function to edit.
We can finally process the Hopedale data set to find the smallest num-
ber of fox pelts produced in any year. As we progress through the file,
we keep the smallest value seen so far in a variable called smallest. That
variable is initially set to the value on the first line, since it’s the small-
est (and only) value seen so far:
Download fileproc/read_smallest.py
import sys
import tsdl
def smallest_value(r):
'''Read and process reader r to find the smallest
value after the TSDL header.'''
line = tsdl.skip_header(r).strip()
print line
# Now line contains the first data value; this is also the
# smallest value found so far, because it is the only one we have seen.
smallest = int(line)
for line in r:
line = line.strip()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 165
value = int(line)
# If we find a smaller value, remember it.
if value < smallest:
smallest = value
return smallest
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
print smallest_value(input_file)
input_file.close()
We could also use this:
smallest = min(smallest, value)
to keep track of the smallest value in the loop in smallest_value.
Data with Missing Values: Fox Pelts
We also have data for colored fox production in Hebron, Labrador:
Download fileproc/hebron.txt
Coloured fox fur production, Hebron, Labrador, 1834-1839
#Source: C. Elton (1942) "Voles, Mice and Lemmings", Oxford Univ. Press
#Table 17, p.265--266
#remark: missing value for 1836
55
262
-
102
178
227
The hyphen indicates that data for the year 1836 is missing. Unfortu-
nately, calling read_smallest on the Hebron data produces this error:
Download fileproc/read_smallest_hebron.cmd
>>> import read_smallest
>>> read_smallest.process_file(open('hebron.txt', 'r'))
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "read_smallest.py", line 15, in process_file
value = int(line)
ValueError: invalid literal for int() with base 10: '-'
The problem is that ’-’ isn’t an integer, so calling int(’-’) fails. This isn’t
an isolated problem. In general, we will often need to skip blank lines,
comments, or lines containing other “nonvalues” in our data. Real data
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 166
sets often contain omissions or contradictions; dealing with them is
just a fact of scientific life.
To fix our code, we must add a check inside the loop that processes
a line only if it contains a real value. In the TSDL data sets, missing
entries are always marked with hyphens, so we just need to check for
that before trying to convert the string we have read to an integer:
Download fileproc/read_smallest_skip.py
import sys
from tsdl import skip_header
def smallest_value_skip(r):
'''Read and process reader r to find the smallest value after
the TSDL header. Skip missing values, which are indicated
with a hyphen.'''
line = skip_header(r).strip()
# Now line contains the first data value; this is also the
# smallest value found so far.
smallest = int(line)
for line in r:
line = line.strip()
# Only process line if it has a valid value.
if line != '-':
value = int(line)
# Process value; if we find a smaller value, remember it.
if value < smallest:
smallest = value
return smallest
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
print smallest_value_skip(input_file)
input_file.close()
Notice that the comparison of value with smallest is nested inside the
check for hyphens. If it were not, then if the line contained a hyphen,
comparing value with smallest would result in an error.
Individual Whitespace-Delimited Data
The file [Hynnd] contains information about lynx pelts in the years
1821–1934. All data values are integers, each line contains many
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 167
values, the values are separated by whitespace, and for reasons best
known to the file’s author, each value ends with a period.
Download fileproc/lynx.txt
Annual Number of Lynx Trapped, MacKenzie River, 1821-1934
#Original Source: Elton, C. and Nicholson, M. (1942)
#"The ten year cycle in numbers of Canadian lynx",
#J. Animal Ecology, Vol. 11, 215--244.
#This is the famous data set which has been listed before in
#various publications:
#Cambell, M.J. and Walker, A.M. (1977) "A survey of statistical work on
#the MacKenzie River series of annual Canadian lynx trappings for the years
#1821-1934 with a new analysis", J.Roy.Statistical Soc. A 140, 432--436.
269. 321. 585. 871. 1475. 2821. 3928. 5943. 4950. 2577. 523. 98.
184. 279. 409. 2285. 2685. 3409. 1824. 409. 151. 45. 68. 213.
546. 1033. 2129. 2536. 957. 361. 377. 225. 360. 731. 1638. 2725.
2871. 2119. 684. 299. 236. 245. 552. 1623. 3311. 6721. 4245. 687.
255. 473. 358. 784. 1594. 1676. 2251. 1426. 756. 299. 201. 229.
469. 736. 2042. 2811. 4431. 2511. 389. 73. 39. 49. 59. 188.
377. 1292. 4031. 3495. 587. 105. 153. 387. 758. 1307. 3465. 6991.
6313. 3794. 1836. 345. 382. 808. 1388. 2713. 3800. 3091. 2985. 3790.
674. 81. 80. 108. 229. 399. 1132. 2432. 3574. 2935. 1537. 529.
485. 662. 1000. 1590. 2657. 3396.
To process this, we must break each line into pieces and strip off the
periods. Our algorithm is the same as it was for the fox pelt data: find
and process the first “real” line in the file, and then process each of the
subsequent lines. However, the notion of “processing a line” needs to be
examined further, because there are many values per line. Our refined
algorithm, shown next, uses nested loops to handle the notion of “for
each line and for each value on that line”:
Find the first line containing real data after the header.
For each piece of data in the current line:
Process that piece.
For each other line of data:
For each piece of data in the current line:
Process that piece.
Once again, we are processing lines in two different places. That is a
strong hint that we should write a helper function to avoid duplicate
code. Rewriting our algorithm, and making it specific to the problem of
finding the largest value, makes this clearer:
Find the first line of real data after the header.
Find the largest value in that line.
For each other line of data:
Find the largest value in that line.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 168
If that value is larger than the previous largest, remember it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 169
The helper function required is one that finds the largest value in a line,
and it must split the line up. The string method split will split around
the whitespace, but we still have to remove the periods at the ends of
values.
We can also simplify our code by initializing largest to -1, because that
value is guaranteed to be smaller than any of the (positive) values in
the file. That way, no matter what the first real value is, it will be larger
than the “previous” value (our -1) and replace it.
Download fileproc/read_lynx_1.py
import sys
def find_largest(line):
'''Return the largest value in line, which is a
whitespace-delimited string of integers.'''
# The largest value seen so far.
largest = -1
for value in line.split():
# Remove the trailing period.
v = int(value[:-1])
# If we find a larger value, remember it.
if v > largest:
largest = v
return largest
We now face the same choice as with skip_header: we can put find_largest
in a module (possibly tsdl), or we can include it in the same file as the
rest of the code. We choose the latter this time, because the code is
specific to this particular data set and problem:
Download fileproc/read_lynx.py
import sys
from tsdl import skip_header
def find_largest(line):
'''Return the largest value in line, which is a
whitespace-delimited string of integers.'''
# The largest value seen so far.
largest = -1
for value in line.split():
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ONE RECORD PER LINE 170
# Remove the trailing period.
v = int(value[:-1])
# If we find a larger value, remember it.
if v > largest:
largest = v
return largest
def process_file(r):
'''Read and process reader r.'''
line = skip_header(r).strip()
# The largest value so far.
largest = find_largest(line)
# Check the rest of the lines for larger values.
for line in r:
large = find_largest(line)
if large > largest:
largest = large
return largest
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
print process_file(input_file)
input_file.close()
Notice how simple the code in process_file looks! This happened only
because we decided to write helper functions. To show you how much
clearer this is, here is the same code without using find_largest as a
helper method:
Download fileproc/read_lynx_expanded.py
import sys
from tsdl import skip_header
def process_file(r):
'''Read and process reader r.'''
line = skip_header(r).strip()
# The largest value seen so far.
largest = -1
for value in line.split():
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
RECORDS WITH MULTIPLE FIELDS 171
# Remove the trailing period.
v = int(value[:-1])
# If we find a larger value, remember it.
if v > largest:
largest = v
# Check the rest of the lines for larger values.
for line in r:
# The largest value seen so far.
large = -1
for value in line.split():
# Remove the trailing period.
v = int(value[:-1])
# If we find a larger value, remember it.
if v > large:
large = v
if large > largest:
largest = large
return largest
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
print process_file(input_file)
input_file.close()
8.2 Records with Multiple Fields
Here is some United States housing data for 1983 and 1984, also taken
from [Hynnd]. The first column is the monthly housing starts (thou-
sands of units), the second is the total construction contracts (millions
of dollars), and the third is the average interest rate for a new home
mortgage (percent):
Download fileproc/housing.dat
91.3 11.358 13
96.3 11.355 12.62
134.6 16.100 12.97
135.8 16.315 12.02
174.9 19.205 12.21
173.2 20.263 11.9
161.6 16.885 12.02
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
RECORDS WITH MULTIPLE FIELDS 172
176.8 19.441 12.01
154.9 17.379 12.08
159.3 16.028 11.8
136 15.401 11.82
108.3 13.518 11.94
109.1 14.023 11.8
130 14.442 11.78
137.5 17.916 11.56
172.7 17.655 11.55
180.7 21.990 11.68
184 20.036 11.61
162.1 19.224 11.91
147.4 19.367 11.89
148.5 16.923 12.03
152.3 18.413 12.27
126.2 16.616 12.27
98.9 14.220 12.05
This data differs from the previous example, because although there
are multiple values per line, in this case each value represents some-
thing different. We want to compare total housing starts and construc-
tion contracts from 1983 to 1984; for the moment, we don’t care about
interest rates.
What if we decide to ask more questions about this data in the future?
Instead of rereading the data from the file, we can store the data for
future use. But how will we store it? We can create twelve lists, one
for each month, or two lists, one for housing starts and one for total
construction contracts, and store the data by column. Another option
is to create a list of lists to keep all the data together. Twelve variables
feels like too many, so let’s store the data by column using two lists.
(A lot of program design is based on what “feels” right. There are no
universal hard-and-fast rules for good design; there are only trade-offs
and consequences.) Using the two lists to store the data, we compare
housing starts and construction contracts from 1983 to 1984:
Download fileproc/housing.py
import sys
def housing(r):
'''Return the difference between the housing starts and
construction contracts in 1983 and in 1984 from reader r.'''
# The monthly housing starts, in thousands of units.
starts = []
# The construction contracts, in millions of dollars.
contracts = []
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
RECORDS WITH MULTIPLE FIELDS 173
# Read the file, populating the lists.
for line in r:
start, contract, rate = line.split()
starts.append(float(start))
contracts.append(float(contract))
return (sum(starts[12:23]) - sum(starts[0:11]),
sum(contracts[12:23]) - sum(contracts[0:11]))
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
print housing(input_file)
input_file.close()
The result is the tuple (55.799999999999955, 16.875000000000028), showing
that both housing starts and construction contracts rose from 1983 to
1984.
This program answered our question, but it could still be improved. Its
first shortcoming is that it throws away the interest rate data; although
we don’t need this right now, someone might in future, so we should
create a third list and store it. The second improvement is to separate
the parsing and processing of the data, that is, to have one function
that reads the data and another that does calculations on it. That way,
we can reuse the parsing code every time we have new questions.
Download fileproc/housing_2.py
import sys
def read_housing_data(r):
'''Read housing data from reader r, returning lists of starts,
contracts, and rates.'''
starts = []
contracts = []
rates = []
for line in r:
start, contract, rate = line.split()
starts.append(float(start))
contracts.append(float(contract))
rates.append(rate)
return (starts, contracts, rates)
def process_housing_data(starts, contracts):
'''Return the difference between the housing starts and
construction contracts in 1983 and in 1984.'''
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
POSITIONAL DATA 174
1
Year Month Day
8 2 7 0 3 1 5 4 3 3 9 3 4 7 9 2 3 4 9 ...
Date Latitude Longitude
Deg. Min. Sec. Deg. Min. Sec.
Figure 8.1: A fixed-width file format
return (sum(starts[12:23]) - sum(starts[0:11]),
sum(contracts[12:23]) - sum(contracts[0:11]))
if __name__ == "__main__":
input_file = open(sys.argv[1], "r")
starts, contracts, rates = read_housing_data(input_file)
print process_housing_data(starts, contracts)
input_file.close()
Many programs go one step further and separate parsing, processing,
and reporting (the printing of results). That way, both the input and
output can be used in other programs without having to be rewritten,
and programs can process data from other sources [Wil05].
8.3 Positional Data
Some file formats don’t use delimiters to separate fields. Instead, each
field is in a fixed location on the line. For example, characters 1–8 might
store the date (with four digits for the year, two for the month, and two
for the day), characters 9–14 and 15–20 the latitude and longitude (as
degrees, minutes, and seconds), and the 24 characters after that the
temperature, humidity, and pressure as decimal numbers, each eight
characters long (see Figure 8.1).
Processing files like this is relatively straightforward, because slicing
strings is easy in Python. Here’s a function that reads a file in the
format just described. Its result is a list of tuples; each tuple is a single
record, and the fields in each tuple are that record’s values.
Download fileproc/fixed_width_1.py
def read_weather_data(r):
'''Read weather data from reader r in fixed-width format.
The fields are:
1 8 YYYYMMDD (date)
9 14 DDMMSS (latitude)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
POSITIONAL DATA 175
15 20 DDMMSS (longitude)
21 26 FF.FFF (temp, deg. C)
27 32 FF.FFF (humidity, %)
33 38 FF.FFF (pressure, kPa)
The result is a list of tuples:
((Yr, Mo, Day), (Deg, Min, Sec), (Deg, Min, Sec), (Temp, Hum, Press))
'''
result = []
for line in r:
year = int(line[0:4])
month = int(line[4:6])
day = int(line[6:8])
lat_deg = int(line[8:10])
lat_min = int(line[10:12])
lat_sec = int(line[12:14])
long_deg = int(line[14:16])
long_min = int(line[16:18])
long_sec = int(line[18:20])
temp = float(line[20:26])
hum = float(line[26:32])
press = float(line[32:38])
result.append(((year, month, day),
(lat_deg, lat_min, lat_sec),
(long_deg, long_min, long_sec),
(temp, hum, press)))
return result
This function does indeed do what we need, but an experienced pro-
grammer would find fault with it. The biggest criticism would be that it
would be very easy to mistype some of those twenty-four slice indices
and wind up taking too much or too little data, like this:
long_min = int(line[16:18]) # missing 18!
long_sec = int(line[19:20])
temp = float(line[20:26])
Also, if the data format ever changes—for example, if this function
needs to read files that used decimal degrees for latitude and longi-
tude—then all the indices after the point of change would need to be
updated.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
POSITIONAL DATA 176
Here’s a more elegant solution:2
Download fileproc/fixed_width_2.py
def read_weather_data(r):
'''Read weather data from reader r in fixed-width format.
The field widths are:
4,2,2 YYYYMMDD (date)
2,2,2 DDMMSS (latitude)
2,2,2 DDMMSS (longitude)
6,6,6 FF.FFF (temp, deg. C; humidity, %; pressure, kPa)
The result is a list of values (not tuples):
(YY, MM, DD, DD, MM, SS, DD, MM, SS, Temp, Hum, Press)'''
fields = ((4, int), (2, int), (2, int), # date
(2, int), (2, int), (2, int), # latitude
(2, int), (2, int), (2, int), # longitude
(6, float), (6, float), (6, float)) # data
result = []
# For each record
for line in r:
start = 0
record = []
# for each field in the record
for (width, target_type) in fields:
# convert the text
text = line[start:start+width]
field = target_type(text)
# add it to the record
record.append(field)
# move on
start += width
# add the completed record to the result
result.append(record)
return result
The basic idea is that each field is a fixed width, so after processing a
field of width W, we just move W characters forward in the string to
find the next field. And since the users of this function don’t want text
back, but integers and floating-point numbers, we store a reference to
the appropriate conversion function right beside each field’s width. An
interesting feature of this example is that the functions are stored in
tuples just like other data.
2. Astute readers will notice that it returns a single tuple per record, rather than a tuple
of tuples. This will be fixed in the exercises.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MULTILINE RECORDS 177
8.4 Multiline Records
Not every data record will fit onto a single line. Here is a file in simplified
Protein Data Bank (PDB) format that describes the arrangements of
atoms in ammonia:
Download fileproc/ammonia.pdb
COMPND AMMONIA
ATOM 1 N 0.257 -0.363 0.000
ATOM 2 H 0.257 0.727 0.000
ATOM 3 H 0.771 -0.727 0.890
ATOM 4 H 0.771 -0.727 -0.890
END
The first line is the name of the molecule. All subsequent lines down to
the one containing END specify the ID, type, and XYZ coordinates of one
of the atoms in the molecule.
Reading this file is straightforward using the tools we have built up in
this chapter. But what if the file contained two or more molecules, like
this:
Download fileproc/multimol.pdb
COMPND AMMONIA
ATOM 1 N 0.257 -0.363 0.000
ATOM 2 H 0.257 0.727 0.000
ATOM 3 H 0.771 -0.727 0.890
ATOM 4 H 0.771 -0.727 -0.890
END
COMPND METHANOL
ATOM 1 C -0.748 -0.015 0.024
ATOM 2 O 0.558 0.420 -0.278
ATOM 3 H -1.293 -0.202 -0.901
ATOM 4 H -1.263 0.754 0.600
ATOM 5 H -0.699 -0.934 0.609
ATOM 6 H 0.716 1.404 0.137
END
As always, we tackle this problem by dividing into smaller ones and
solving each of those in turn. Our first algorithm is as follows:
while there are more molecules in the file:
read a molecule from the file
append it to the list of molecules read so far
Simple, except the only way to tell whether there is another molecule
left in the file is to try to read it. Our modified algorithm is as follows:
reading = True
while reading:
try to read a molecule from the file
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MULTILINE RECORDS 178
if there is one:
append it to the list of molecules read so far
else: # nothing left
reading = False
In Python, this is as follows:
Download fileproc/multimol.py
def read_all_molecules(r):
'''Read zero or more molecules from reader r,
returning a list of the molecules read.'''
result = []
reading = True
while reading:
molecule = read_molecule(r)
if molecule:
result.append(molecule)
else:
reading = False
return result
The work of actually reading a single molecule has been put in a func-
tion of its own that must return some false value (such as None) if it
can’t find another molecule in the file. This function checks the first line
it tries to read to see whether there is actually any data left in the file. If
not, it returns immediately to tell read_all_molecules that the end of the
file has been reached. Otherwise, it pulls the name of the molecule out
of the first line and then reads the molecule’s atoms one at a time down
to the END line:
Download fileproc/multimol_2.py
def read_molecule(r):
'''Read a single molecule from reader r and return it,
or return None to signal end of file.'''
# If there isn't another line, we're at the end of the file.
line = r.readline()
if not line:
return None
# Name of the molecule: "COMPND name"
key, name = line.split()
# Other lines are either "END" or "ATOM num type x y z"
molecule = [name]
reading = True
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LOOKING AHEAD 179
while reading:
line = r.readline()
if line.startswith('END'):
reading = False
else:
key, num, type, x, y, z = line.split()
molecule.append((type, x, y, z))
return molecule
Notice that this function uses exactly the same trick to spot the END
marking the end of a single molecule as the first function used to spot
the end of file.
8.5 Looking Ahead
Let’s add one final complication. Suppose that molecules didn’t have
END markers but instead just a COMPND line followed by one or more
ATOM lines. How would we read multiple molecules from a single file in
that case?
At first glance, it doesn’t seem much different from the problem we
just solved: read_molecule could extract the molecule’s name from the
COMPND line and then read ATOM lines until it got either an empty
string signaling the end of the file or another COMPND line signaling
the start of the next molecule. But once it has read that COMPND line,
the line isn’t available for the next call to read_molecule, so how can we
get the name of the second molecule (and all the ones following it)?
To solve this problem, our functions must always “look ahead” one line.
Let’s start with the function that reads multiple molecules:
Download fileproc/lookahead.py
def read_all_molecules(r):
'''Read zero or more molecules from reader r,
returning a list of the molecules read.'''
result = []
line = r.readline()
while line:
molecule, line = read_molecule(r, line)
result.append(molecule)
return result
This function begins by reading the first line of the file. Provided that
line is not the empty string (that is, the file being read is not empty),
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LOOKING AHEAD 180
'AMMONIA'
molecule
0 1 2 3 4
0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3
'N'
0.257
-0.363
0.000
'H'
0.257
0.727
0.000
'H'
0.771
-0.727
0.890
'H'
0.771
-0.727
-0.890
Figure 8.2: A PDB file
it passes both the stream to read from and the line into read_molecule,
which is supposed to return two things: the next molecule in the file
and the first line immediately after the end of that molecule (or an
empty string if the end of file has been reached).
This simple description is enough to get us started writing the read_
molecule function. The first thing it has to do is check that line is actu-
ally the start of a molecule. It then reads lines from stream one at a time,
looking for one of three situations:
• The end of file, which signals the end of both the current molecule
and the file
• Another COMPND line, which signals the end of this molecule and
the start of the next one
• An ATOM, which is to be added to the current molecule
The most important thing is that when this function returns, it returns
both the molecule and the next line so that its caller can keep process-
ing. The result is probably the most complicated function we have seen
so far, but understanding the idea behind it will help you understand
how it works.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WRITING TO FILES 181
'AMMONIA'
molecule0
0 ...
next0
'COMPND METHANOL\n'
'METHANOL'
molecule1
0 ...
next1
''
Figure 8.3: Looking ahead
Download fileproc/lookahead_2.py
def read_molecule(r, line):
'''Read a molecule from reader r. The variable 'line'
is the first line of the molecule to be read; the result is
the molecule, and the first line after it (or the empty string
if the end of file has been reached).'''
fields = line.split()
molecule = [fields[1]]
line = r.readline()
while line and not line.startswith('COMPND'):
fields = line.split()
key, num, type, x, y, z = fields
molecule.append((type, x, y, z))
line = r.readline()
return molecule, line
8.6 Writing to Files
In addition to extracting data from files, there are times when we’d like
to modify or create files using Python. For example, when processing
data from a file, we may want to create a new file containing the results.
We may also want to take two data files and merge them into a single
file or split a single file into different files.
To open a file, we provide the filename and one of three modes ’r’, ’w’, or
’a’, which stand for read, write, and append, respectively. If the mode
isn’t provided (that is, we call open with only one argument), then the
default mode is ’r’.
To create a new file or to replace the contents of an existing file, we use
the write mode. If the filename does not exist already, then a new file
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
WRITING TO FILES 182
is created; otherwise, the file contents are replaced. For example, let’s
put “Computer Science” in the file test.txt:
Download fileproc/write.py
output_file = open("test.txt", "w")
output_file.write("Computer Science")
output_file.close()
Rather than replacing the file contents, we can also add to a file using
the append mode. When we write to a file that is opened in append
mode, the data we write is added to the end of the file, and the current
file contents are not overwritten. For example, to add to our previous
file test.txt, we can append the line “Software Engineering”:
Download fileproc/append.py
output_file = open("test.txt", "a")
ouput_file.write("Software Engineering")
output_file.close()
In each of the previous examples, we called write only once, but we
would typically call it multiple times.
The next example is more complex, and it involves both reading from
and writing to a file. Our input file contains two numbers per line sep-
arated by a space. The output file will contain three numbers: the two
from the input file and their sum (all separated by spaces):
Download fileproc/sum.py
def sum(input_file, output_filename):
'''Reads the data from open file descriptor input_file, which contains
two floats per line separated by a space. For each line from
input_file, a line is written to the file named output_filename
containing the two floats from the corresponding line of input_file
plus a space and the sum of the two floats.'''
output_file = open(output_filename, 'w')
for line in input_file:
operands = line.split()
print 'operands', operands
sum = float(operands[0]) + float(operands[1])
new_line = line.rstrip() + ' ' + str(sum) + '\n'
output_file.write(new_line)
output_file.close()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 183
8.7 Summary
In this chapter, we learned the following:
• Data stored in files is usually formatted in one of a small num-
ber of ways, from value per line to multiline records with explicit
end-of-record markers. Each format can be processed in a stereo-
typical way.
• Data processing programs should be broken into input, process-
ing, and output stages so that each can be reused independently.
• Programs that take filenames as command-line arguments are
more flexible, and therefore more useful, than ones in which these
names are hard-coded.
• Files can be read (content retrieved), written to (content replaced),
and added to (new content appended). When a file is opened in
writing mode and it does not exist, a new file is created.
• Data files come in many different formats, so custom code is often
required, but we can reuse as much as possible by writing helper
functions.
• To make the functions usable by different types of readers, the
reader (file, web page or input stream) is opened outside the func-
tion, passed as an argument to the function, and then closed out-
side the function.
8.8 Exercises
Here are some exercises for you to try on your own:
1. All of the file-reading functions we have seen in this chapter read
forward through the file from the first character or line to the
last. How would you write a function that would read backward
through a file?
2. The file reader we developed in Section 8.3, Positional Data, on
page 174 returns a single tuple for each entire record. Modify it so
that if the field format is specified like this:
fields = (
((4, int), (2, int), (2, int)), # date
((2, int), (2, int), (2, int)), # latitude
((2, int), (2, int), (2, int)), # longitude
((6, float), (6, float), (6, float)) # data
)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 184
then each record will be converted into a tuple of tuples, like this:
(
(2007, 5, 3), # date
(24, 17, 37), # latitude
(38, 56, 5), # longitude
(100.0, 121.3, 16.37) # data
)
Explain why a tuple of tuples would make life easier for the pro-
grams that are using this function.
3. The format descriptor used by the fixed-width file reader of Sec-
tion 8.3, Positional Data, on page 174, specifies the width of each
field. However, formats are sometimes specified in terms of start-
ing positions. For example, look at Figure 8.1, on page 174; the
fixed-width file format could be specified as this:
fields = (
(0, int), (4, int), (6, int), # date
(8, int), (10, int), (12, int), # latitude
(14, int), (16, int), (18, int), # longitude
(20, float), (26, float), (32, float) # data
)
Write a function that takes a format specified this way, converts
it into a width-based format, and then reads data from the spec-
ified file. (Functions like this are called wrapper functions, since
they are “wrapped around” existing code so that it can be used in
different ways.)
4. Modify the file reader in read_smallest_skip.py of Section 8.1, Skip-
ping the Header, on page 162 so that can handle files with no
data.
5. Modify the file reader in read_smallest_skip.py of Section 8.1, Skip-
ping the Header, on page 162 so that it uses a continue inside the
loop instead of an if. Which form do you find easier to read?
6. Modify the PDB file reader of Section 8.4, Multiline Records, on
page 177 so that it ignores blank lines and comment lines in PDB
files. A blank line is one that contains only space and tab char-
acters (that is, one that looks empty when viewed). A comment is
any line beginning with the keyword CMNT.
7. Modify the PDB file reader to check that the serial numbers on
atoms start at 1 and increase by 1. What should the modified
function do if it finds a file that doesn’t obey this rule?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 9
Sets and DictionariesSo far, the only way we have seen to store multiple values is to put
them in a sequence, such as a list or a tuple. In this chapter, we will
investigate two other kinds of collections called sets and dictionaries.
These allow us to create programs that are simpler and more efficient
than those we could write using sequences alone.
9.1 Sets
A set is an unordered collection of distinct items. Unordered means that
items are not stored in any particular order. Something is either in the
set or not, but there’s no notion of it being the first, second, or last
item. Distinct means that any item appears in a set at most once; in
other words, there are no duplicates.
Sets are fundamental to mathematics and are built in to modern ver-
sions of Python. To create a new empty set, simply type set(). To create
a set with values already in it, type set((2, 3, 5)). It is important to notice
that the initial values are specified in a single argument, which is a
tuple. We could instead use a list, like set([2, 3, 5]), but cannot pass the
values one by one, like set(2, 3, 5).
It’s equally important to understand that the order of the values doesn’t
matter, nor does the number of times each value is entered. The expres-
sions set((3, 5, 2)) and set((2, 3, 5, 5, 2, 3)) create exactly the same result: a
set containing just three values (see Figure 9.1, on the following page).
Set Operations
In mathematics, set operations include union, intersection, add, and
remove. In Python, these are implemented as methods (for a complete
Prepared exclusively for Trieu Nguyen
SETS 186
2
�5
set([3, 5, 2]) set([2, 3, 5, 5, 2, 3])
2
35
Figure 9.1: Creating a set
list, see Figure 9.2, on the next page). Note that most operations cre-
ate a new set: only add, remove, and clear modify the current set. The
following program shows these methods in action:
Download setdict/setexamples.cmd
>>> ten = set(range(10))
>>> lows = set([0, 1, 2, 3, 4])
>>> odds = set([1, 3, 5, 7, 9])
>>> lows.add(9)
>>> lows
set([0, 1, 2, 3, 4, 9])
>>> lows.difference(odds)
set([0, 2, 4])
>>> lows.intersection(odds)
set([1, 3, 9])
>>> lows.issubset(ten)
True
>>> lows.issuperset(odds)
False
>>> lows.remove(0)
>>> lows
set([1, 2, 3, 4, 9])
>>> lows.symmetric_difference(odds)
set([2, 4, 5, 7])
>>> lows.union(odds)
set([1, 2, 3, 4, 5, 7, 9])
>>> lows.clear()
>>> lows
set([])
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SETS 187
Method Purpose Example Result
add Adds an element to a set lows.add(9) None
clear Removes all elements from a set
lows.clear() None
difference Creates a set with elements from one set, but not the other
lows.difference(odds) set([0, 2, 4]])
intersection Creates a set with elements that are in both sets
lows.intersection(odds) set([1, 3]])
issubset Asks are all of one set’s elements contained in another?
lows.issubset(ten) True
issuperset Asks does one set contain all of another’s elements?
lows.issuperset(odds) False
remove Removes an element from a set
lows.remove(0) None
symmetric_difference Creates a set with elements that are in exactly one set
lows.symmetric_difference(odds) set([0, 2, 4, 5, 7, 9]])
union Creates a set with elements that are in either set
lows.union(odds) set([0, 1, 2, 3, 4, 5, 7, 9]])
Figure 9.2: Set operations, where ten = set(range(10)), lows = set([0, 1,
2, 3, 4]), and odds = set([1, 3, 5, 7, 9])
Many set methods can also be written using operators. If acids and bases
are two sets, for example, then acids | bases creates a new set containing
their union (that is, all the elements from both acids and bases), while
acids <= bases tests whether all the values in acids are also in bases. All
the operators that sets support are listed in Figure 9.3, on the following
page.
Arctic Birds
To see how sets are used, suppose we have several files recording obser-
vations of birds in the Canadian Arctic and we want to know which
species we have seen. Each file is formatted like this:
Download setdict/birdwatching.txt
canada goose
canada goose
long-tailed jaeger
canada goose
snow goose
canada goose
canada goose
northern fulmar
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SETS 188
Method Call Operator
set1.difference(set2) set1 - set2
set1.intersection(set2) set1 & set2
set1.issubset(set2) set1 <= set2
set1.issuperset(set2) set1 >= set2
set1.union(set2) set1 | set2
set1.symmetric_difference(set2) set1∧set2
Figure 9.3: Set operations
Here’s our program:
Download setdict/birdwatching.py
import sys
# Find the different bird types observed.
birds = set()
for filename in sys.argv[1:]:
infile = open(filename, 'r')
for line in infile:
name = line.strip()
birds.add(name)
infile.close()
# Print the birds.
for b in birds:
print b
The first statement after the import creates a new empty set and assigns
it to the variable birds. The loop then reads the names of birds from each
of the input files specified on the command line. Any whitespace before
or after the bird’s name is stripped away by line.strip(); the program then
uses birds.add(name) to make sure the new name is in the set. If the
name is not already there, set.add puts it in the set; otherwise, the set
isn’t changed.
Once all the names have been read, the program loops over the values
in the set to print them. This works exactly like a loop over the val-
ues in a list, except that the order in which values are encountered is
arbitrary: there is no guarantee that they will come out in the order
in which they were added, in alphabetical order, in order by length, or
anything else.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SETS 189
A-H I-O P-Z
Marie Sklodowska
A-H I-O P-Z
Marie Curie (nee Sklodowska)
Should be filed under A-H
Figure 9.4: Moving items in a filing cabinet
Set Storage
Sets are stored in a data structure called a hash table. Each time an
item is added to a set, Python calculates a hash code for the item,
which is an integer that is guaranteed to be the same for items with
equal values:
Download setdict/hashcode.cmd
>>> help(hash)
Help on built-in function hash in module __builtin__:
hash(...)
hash(object) -> integer
Return a hash value for the object. Two objects with the same
value have the same hash value. The reverse is not necessarily
true, but likely.
>>> hash(123)
123
>>> hash('123') # a string
1911471187
To see whether a value is in a set, Python simply recalculates the hash
code for that value and checks the corresponding location in the hash
table. This is a lot faster than searching the whole set every time a value
needs to be looked up. However, this scheme works only if an item’s
hash code cannot change after it has been added to the hash table.
To see why, consider what would happen if you filed patient records in
alphabetical order but then went back and changed someone’s name
(see Figure 9.4). Their record would now be out of place, so if someone
else tried to look it up, they wouldn’t find it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 190
Because hash codes are computed from values, if a list’s contents are
changed, the hash code for that list will change. For this reason, Python
allows sets to contain only immutable values (see Section 5.2, Modifying
Lists, on page 85). Again, Booleans, numbers, strings, and tuples are
allowed, but lists are not:
Download setdict/hashcode_seq.cmd
>>> hash((1, 2, 3))
-378539185
>>> hash([1, 2, 3])
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: list objects are unhashable
This is actually one of the reasons tuples were invented: they allow
multipart values like (’Albert’, ’Einstein’) to be added to sets.
But this means that we can’t store a set of sets. Sets themselves can’t
be immutable, since we need to add and remove values, so a set can’t
contain another one. To solve this problem, Python has another data
type called a frozen set. As the name implies, frozen sets are sets whose
contents can’t be changed, just as tuples are lists whose values can’t be
modified. An empty frozen set (which isn’t particularly useful) is created
using frozenset(); to create a frozen set that contains some values, use
frozenset(values), where values is a list, tuple, set, or other collection.
9.2 Dictionaries
Back to bird watching. Suppose we want to know how often each kind
of bird was seen. Our first attempt uses a list of pairs, each a list storing
the name of a bird and the number of times it has been seen so far:
Download setdict/listfordict.py
import sys
# Find all the birds.
birds = []
for filename in sys.argv[1:]:
infile = open(filename, 'r')
# For each bird, find its entry and increment the count.
for line in infile:
name = line.strip()
found = False
for entry in birds:
if entry[0] == name:
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 191
entry[1] += 1
found = True
if not found:
birds.append([name, 1])
infile.close()
# Print.
for (name, count) in birds:
print name, count
This gives the right answer, but there are two things wrong with it. The
first is that it is complex. The more nested loops our programs contain,
the harder they are to understand, fix, and extend. The second is that it
is inefficient. Suppose we were interested in beetles instead of birds and
that we had millions of observations of tens of thousands of species.
Scanning the list of names each time we want to add one new observa-
tion would take a long, long time, even on a fast computer (a topic we
will return to in Chapter 11, Searching and Sorting, on page 214).
Can we use a set to solve both problems at once? Sets can look up
values in a single step; why not combine each bird’s name and the
number of times it has been seen into a two-valued tuple and put those
tuples in a set?
The problem with this idea is that we can look for values only if we
know what those values are. In this case, we won’t. We will know only
the name of the species, not how many times it has already been seen.
The right answer is to use another data structure called a dictionary.
Also known as a map, a dictionary is an unordered mutable collec-
tion of key/value pairs (see Figure 9.5, on the following page). In plain
English, dictionaries are like phone books. They associate a value (such
as a phone number) with a key (like a name). The keys form a set. Any
particular key can appear at most once in a dictionary, and like the ele-
ments in sets, keys must be immutable (though the values associated
with them do not have to be).
Dictionaries are created by putting key/value pairs inside braces. To
get the value associated with a key, we put the key in square brackets:
Download setdict/simpledict.cmd
>>> birds = {'canada goose' : 3, 'northern fulmar' : 1}
>>> birds
{'canada goose' : 3, 'northern fulmar' : 1}
>>> birds['northern fulmar']
1
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 192
'canada goose'
'northern fulmar'
3 1
Figure 9.5: Structure of a dictionary
As you’d expect, the empty dictionary is written {}. Indexing a dictionary
with a key it doesn’t contain produces an error, just like an out-of-range
index for a list:
Download setdict/badkey.cmd
>>> birds = {'canada goose' : 3, 'northern fulmar' : 1}
>>> birds['canada goose']
3
>>> birds['long-tailed jaeger']
Traceback (most recent call last):
File "<stdin>", line 1, in ?
KeyError: 'long-tailed jaeger'
To test whether a key is in a dictionary, use k in d:
Download setdict/dictin.cmd
>>> birds = {'eagle' : 999, 'snow goose' : 33}
>>> if 'eagle' in birds:
... print 'eagles have been seen'
...
eagles have been seen
>>> del birds['eagle']
>>> if 'eagle' in birds:
... print 'oops: why are eagles still there?'
...
Updating and Membership
Updating dictionaries is easy. Just assign a value to a key. If the key is
already in the dictionary, this changes the value associated with it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 193
If the key was not present, it is added, along with the value:
Download setdict/dictupdate.cmd
>>> birds = {}
>>> birds['snow goose'] = 33
>>> birds['eagle'] = 999 # oops
>>> birds
{'eagle' : 999, 'snow goose' : 33}
>>> birds['eagle'] = 9
>>> birds
{'eagle' : 9, 'snow goose' : 33}
To remove an entry from a dictionary, use del d[k], where d is the dictio-
nary and k is the key being removed. Only entries that are present can
be removed; trying to remove one that isn’t there causes an error:
Download setdict/dictdel.cmd
>>> birds = {'snow goose' : 33, 'eagle' : 9}
>>> del birds['snow goose']
>>> birds
{'eagle' : 9}
>>> del birds['gannet']
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
KeyError: 'gannet'
Loops
Since dictionaries are collections, we’re going to want to loop over their
contents. We do this with for key in somedict, which assigns each of the
keys in the dictionary to the loop variable in turn:
Download setdict/loop.cmd
>>> birds = {'canada goose' : 183, 'long-tailed jaeger' : 71,
'snow goose' : 63, 'northern fulmar', 1}
>>> for x in birds:
... print x, birds[x]
...
'northern fulmar' 1
'long-tailed jaeger' 71
'canada goose' 183
'snow goose' 63
As with set elements, Python loops over the entries in the dictionary in
an arbitrary order. There is no guarantee that they will be seen alpha-
betically or in the order they were added to the dictionary.
Notice, by the way, that looping over dictionaries is slightly different
from looping over lists. When Python loops over a list, the values in the
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 194
list are assigned to the loop variable. When it loops over a dictionary,
on the other hand, it assigns the keys. Python’s designers chose to do
this because
• looping over the indices of a list isn’t very interesting, since the
program would always get the sequence 0, 1, 2, ...; and
• it’s a lot easier to go from a dictionary key to the associated value
than it is to take the value and find the associated key.
Dictionary Methods
Like lists, tuples, and sets, dictionaries are objects. Their methods are
described in Figure 9.6, on the next page, while the following program
shows how they are used:
Download setdict/dictmeth.py
scientists = {'Newton' : 1642, 'Darwin' : 1809, 'Turing' : 1912}
print 'keys:', scientists.keys()
print 'values:', scientists.values()
print 'items:', scientists.items()
print 'get:', scientists.get('Curie', 1867)
temp = {'Curie' : 1867, 'Hopper' : 1906, 'Franklin' : 1920}
scientists.update(temp)
print 'after update:', scientists
scientists.clear()
print 'after clear:', scientists
As you can see from its output (shown next), the keys and values meth-
ods return the dictionary’s keys and values, respectively, while items
returns a list of (key, value) pairs. get returns the value associated with
a key or returns some user-specified value if the key isn’t in the dic-
tionary; we’ll see a use for this shortly. Finally, update copies keys and
values from one dictionary into another, and clear erases the dictio-
nary’s contents.
Download setdict/dictmeth.out
keys: ['Turing', 'Newton', 'Darwin']
values: [1912, 1642, 1809]
items: [('Turing', 1912), ('Newton', 1642), ('Darwin', 1809)]
get: 1867
after update: {'Curie': 1867, 'Darwin': 1809, 'Franklin': 1920,
'Turing': 1912, 'Newton': 1642, 'Hopper': 1906}
after clear: {}
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 195
Method Purpose Example Result
clear Empties the dic-
tionary.
d.clear() Returns None, but
d is now empty.
get Returns the value
associated with a
key, or a default
value if the key is
not present.
d.get(’x’, 99) Returns d[’x’] if "x"
is in d, or 99 if it
is not.
keys Returns the dic-
tionary’s keys as
a list. Entries are
guaranteed to be
unique.
birthday.keys() [’Turing’, ’Newton’,
’Darwin’]
items Returns a list of
(key, value) pairs.
birthday.items() [(’Turing’, 1912),
(’Newton’, 1642),
(’Darwin’, 1809)]
values Returns the dic-
tionary’s values as
a list. Entries may
or may not be
unique.
birthday.values() [1912, 1642, 1809]
update Updates the
dictionary with
the contents of
another.
See the example
on the preceding
page.
Figure 9.6: Dictionary methods
One common use of items is to loop over the keys and values in a dic-
tionary together:
for (key, value) in dictionary.items():
...do something with the key and value...
This is inefficient for large dictionaries, since items actually constructs
a list of (key, value) pairs. A similar method called iteritems hands these
pairs back one by one on demand:
for (key, value) in dictionary.iteritems():
...do something with the key and value...
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DICTIONARIES 196
Dictionary Example
Back to birdwatching once again. We want to count the number of times
each species has been seen. To do this, we create a dictionary that is
initially empty. Each time we read an observation from a file, we check
to see whether we have seen that bird before, that is, whether the bird’s
name is already a key in our dictionary. If it is, we add one to its count.
If it isn’t, we add the name to the dictionary with the value 1. Here is
the program that does this:
Download setdict/countbirds1.py
import sys
# Count all the birds.
count = {}
for filename in sys.argv[1:]:
infile = open(filename, 'r')
for line in infile:
name = line.strip()
if name in count:
count[name] = count[name] + 1
else:
count[name] = 1
infile.close()
# Print.
for b in count:
print b, count[b]
We can shorten this program a bit using the method dict.get. This
returns either the value associated with a key or some default value
that we provide. In this case, we get either the number of times we’ve
already seen a species or zero, add one to whichever value the method
returns, and store that back in the dictionary:
Download setdict/countbirds2.py
import sys
# Count all the birds.
count = {}
for filename in sys.argv[1:]:
infile = open(filename, 'r')
for line in infile:
name = line.strip()
count[name] = count.get(name, 0) + 1
infile.close()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
INVERTING A DICTIONARY 197
# Print.
keys = count.keys()
keys.sort()
for b in keys:
print b, count[b]
Using dict.get saves three lines, but some programmers find it harder
to understand at a glance.
This program contains another innovation as well. Instead of printing
the birds’ names in whatever order the dictionary chose to use, we get
the dictionary’s keys as a list, sort that list alphabetically, and then
loop over that. This way, the entries appear in a sensible order. We can
make this more readable by using Python’s built-in sorted function:
Download setdict/countbirds3.py
import sys
# Count all the birds.
count = {}
for filename in sys.argv[1:]:
infile = open(filename, 'r')
for line in infile:
name = line.strip()
count[name] = count.get(name, 0) + 1
infile.close()
# Invert the dictionary.
freq = {}
for (name, times) in count.items():
if times in freq:
freq[times].append(name)
else:
freq[times] = [name]
# Print.
for key in sorted(freq):
print key
for name in freq[key]:
print ' ', name
9.3 Inverting a Dictionary
We might want to print the birds in another order—in order of fre-
quency, for example. To do this, we need to invert the dictionary; that
is, use the values as keys and the keys as values. This is a little trickier
than it first appears. There’s no guarantee that the values are unique,
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 198
so we have to handle collisions. For example, if we invert the dictionary
{’a’:1, ’b’:1, ’c’:1}, the key will be 1, but it’s not clear what the value
would be.
The solution is to use some sort of collection, such as a list, to store
the inverted dictionary’s values. If we go this route, the inverse of the
dictionary shown earlier would be {1:[’a’,’b’,’c’]}. Here’s a program to do
what we want:
Download setdict/countbirds4.py
import sys
# Count all the birds.
count = {}
for filename in sys.argv[1:]:
infile = open(filename, 'r')
for line in infile:
name = line.strip()
count[name] = count.get(name, 0) + 1
infile.close()
# Invert the dictionary.
freq = {}
for (name, times) in count.items():
if times in freq:
freq[times].append(name)
else:
freq[times] = [name]
# Print.
for key in sorted(freq):
print key
for name in freq[key]:
print ' ', name
The exercises will ask you to break this into functions to make it more
readable.
9.4 Summary
In this chapter, we learned the following:
• Sets are used in Python to store unordered collections of unique
values. They support the same operations as sets in mathematics.
• Sets are stored in hash tables to make lookup efficient. For this
to work, each item must have a hash code that does not change
during its lifetime. This means that mutable values like lists and
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 199
sets cannot themselves be stored in sets. Instead, programs must
use tuples and frozen sets.
• Dictionaries are used to store unordered collections of key/value
pairs. They are also stored using hash tables for efficiency’s sake
and also require keys to be immutable.
• Looking things up in sets and dictionaries is much faster than
searching through lists. If you have a program that is doing the
latter, consider changing your choice of data structures.
9.5 Exercises
Here are some exercises for you to try on your own:
1. Write a function called find_dups that takes a list of integers as its
input argument and returns a set of those integers that occur two
or more times in the list.
2. Python’s set objects have a method called pop that removes and
returns an arbitrary element from the set. If a set gerbils con-
tains five cuddly little animals, for example, calling gerbils.pop()
five times will return those animals one by one, leaving the set
empty at the end. Use this to write a function called mating_pairs
that takes two equal-sized sets called males and females as input
and returns a set of pairs; each pair must be a tuple containing
one male and one female. (The elements of males and females may
be strings containing gerbil names or gerbil ID numbers—your
function must work with both.)
3. The PDB file format is often used to store information about mol-
ecules. A PDB file may contain zero or more lines that begin with
the word AUTHOR (which may be in uppercase, lowercase, or mixed
case), followed by spaces or tabs, followed by the name of the per-
son who created the file. Write a function that takes a list of file-
names as an input argument and returns the set of all author
names found in those files.
4. Draw a memory model of the data structure produced by the fol-
lowing statement:
vowels = set(frozenset('aeiou'))
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 200
5. The keys in a dictionary are guaranteed to be unique, but the
values are not. Write a function called count_values that takes a
single dictionary as an argument and returns the number of dis-
tinct values it contains. Given the input {’red’:1, ’green’:1, ’blue’:2},
for example, it would return 2.
6. After doing a series of experiments, you have compiled a dictio-
nary showing the probability of detecting certain kinds of sub-
atomic particles. The particles’ names are the dictionary’s keys,
and the probabilities are the values: {’neutron’:0.55, ’proton’:0.21,
’meson’:0.03, ’muon’:0.07, ’neutrino’:0.14}. Write a function that takes
a single dictionary of this kind as input and returns the parti-
cle that is least likely to be observed. Given the dictionary shown
earlier, for example, the function would return ’meson’.
7. Write a function called count_duplicates that takes a dictionary as
an argument and returns the number of values that appear two
or more times.
8. Write a function called fetch_and_set that takes a dictionary and
two arbitrary values, key and new_value, as arguments. If key is
already in the dictionary, this function returns the value associ-
ated with it while replacing that value with new_value. If the key is
not present, the function raises a KeyError exception with the error
message ’Unable to replace value for nonexistent key’.
9. A balanced color is one whose red, green, and blue values add up
to 1.0. Write a function called is_balanced that takes a dictionary
whose keys are ’R’, ’G’, and ’B’ as its input and returns True if they
represent a balanced color.
10. Write a function called dict_intersect that takes two dictionaries
as arguments and returns a dictionary that contains only the
key/value pairs found in both of the original dictionaries.
11. Programmers sometimes use a dictionary of dictionaries as a sim-
ple database. For example, to keep track of information about
famous scientists, they might use something like this:
Download setdict/dict_of_dicts.py
{
'jgoodall' : {'surname' : 'Goodall',
'forename' : 'Jane',
'born' : 1934,
'died' : None,
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 201
'notes' : 'primate researcher'
'author' : ['In the Shadow of Man', 'The Chimpanzees of Gombe']},
'rfranklin' : {'surname' : 'Franklin',
'forename' : 'Rosalind',
'born' : 1920,
'died' : 1957,
'notes' : 'contributed to discovery of DNA'},
'rcarson' : {'surname' : 'Carson',
'forename' : 'Rachel',
'born' : 1907,
'died' : 1964,
'notes' : 'raised awareness of effects of DDT',
'author' : ['Silent Spring']}
}
Write a function called db_headings that returns the set of keys
used in any of the inner dictionaries. In this example, the func-
tion should return set(’author’, ’forename’, ’surname’, ’notes’, ’born’,
’died’).
12. Write another function called db_consistent that takes a dictionary
of dictionaries in the format described in the previous question
and returns True if and only if every one of the inner dictionaries
has exactly the same keys. (This function would return False for
the previous example, since Rosalind Franklin’s entry does not
contain the ’author’ key.)
13. Make the final bird-counting program in Section 9.3, Inverting a
Dictionary, on page 197, more readable by breaking it into several
functions.
14. A sparse vector is a vector whose entries are almost all zero, like [1,
0, 0, 0, 0, 0, 3, 0, 0, 0]. Storing all those zeroes in a list wastes memory,
so programmers often use dictionaries instead to keep track of just
the nonzero entries. For example, the vector shown earlier would
be represented as {0:1, 6:3}, since the vector it is meant to represent
has the value 1 at index 0 and the value 3 at index 6.
a) Write a function called sparse_add that takes two sparse vec-
tors stored as dictionaries and returns a new dictionary rep-
resenting their sum.1
1. The sum of two vectors is just the element-wise sum of their elements. For example,
the sum of [1, 2, 3] and [4, 5, 6] is [5, 7, 9].
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 202
b) Write another function called sparse_dot that calculates the
dot product of two sparse vectors.2
c) Your boss has asked you to write a function called sparse_len
that will return the length of a sparse vector (just as Python’s
len returns the length of a list). What do you need to ask her
before you can start writing it?
2. The dot product of two vectors is the sum of the products of corresponding elements.
For example, the dot product of [1, 2, 3] and [4, 5, 6] is 4+10+18, or 32.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 10
AlgorithmsFor the past few chapters, you have used built-in functions such as
max and sort on lists. This chapter will examine how they work and
thus introduce a systematic way of solving problems when there is no
function or method that does exactly what you want.
An algorithm is a set of steps that accomplishes a task, such as syn-
thesizing caffeine. Although programs require algorithms to be written
in programming languages like Python, it is easier to discuss and plan
an algorithm using a mixture of a human language like English and a
little mathematics.
Our algorithm-writing technique is called top-down design. You start
by describing your solution in English and then mark the phrases that
correspond directly to Python statements. Those that do not correspond
should be rewritten in more detail in English, until everything in your
description can be written in Python.
Top-down design sounds easy, but doing it requires a little practice. As
we work through the examples in the rest of this chapter, we will see
why our initial attempts at translation will have implementations that
look reasonable but often have bugs. This is common in many fields.
In mathematics, for example, the first versions of “proofs” often han-
dle common cases well but fail for odd cases [Lak76]. Mathematicians
deal with this by looking for counter examples, and programmers (good
programmers, at least) deal with it by testing their code as they write it.
Most important, this chapter focuses only on top-down design; we have
skipped the discussion of testing to make sure our code works. In fact,
the first versions we wrote had minor bugs in them, and we found them
Prepared exclusively for Trieu Nguyen
SEARCHING 204
only by doing thorough testing. We will talk more about this in Chap-
ter 12, Construction, on page 237.
10.1 Searching
To start, suppose we have data showing the number of humpback
whales sighted off the coast of British Columbia over the past ten years:
809 834 477 478 307 122 96 102 324 476
We want to know how changes in fishing practices have impacted the
whales’ numbers. Our first question is, which year had the lowest num-
ber of sightings during those years? This code tells us just that:
Download alg/search1.cmd
>>> counts = [809, 834, 477, 478, 307, 122, 96, 102, 324, 476]
>>> min(counts)
96
If we want to know in which year the population bottomed out, we can
use index to find the position of the minimum:
Download alg/search2.cmd
>>> counts = [809, 834, 477, 478, 307, 122, 96, 102, 324, 476]
>>> low = min(counts)
>>> min_index = counts.index(low)
>>> print min_index
6
or more succinctly:
Download alg/search3.cmd
>>> counts = [809, 834, 477, 478, 307, 122, 96, 102, 324, 476]
>>> counts.index(min(counts))
6
Although this does the job, it seems a little inefficient to search through
the data to find the minimum value and then search again to find that
value’s index. Intuitively speaking, there ought to be a way to find both
in just one pass.
Now what if we want to find the indices of the two smallest values?
Lists don’t have a method to do this directly, so we will have to write a
function ourselves.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SEARCHING 205
There are at least three ways we could do this, each of which will be
subjected to top-down design:
• Find, remove, find. Find the index of the minimum, remove that
element from the list, and find the index of the new minimum
element in the list.
• Sort, identify minimums, get indices. Sort the list, get the two small-
est numbers, and then find their indices in the original list.
• Walk through the list. Examine each value in the list in order, keep
track of the two smallest values found so far, and update these
values when a new smaller value is found.
At the end of all three, we’ll return a tuple of the two smallest indices.
Find, Remove, Find
Here is the algorithm again, rewritten with one instruction per line:
Download alg/find_remove_find1.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
find the index of the minimum element in L
remove that element from the list
find the index of the new minimum element in the list
return the two indices
To refine “find the index of the minimum element in L,” we skim the
help for list and find that there are no methods that do exactly that.
We’ll refine it:
Download alg/find_remove_find2.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
get the minimum element in L
find the index of that minimum element
remove that element from the list
find the index of the new minimum element in the list
return the two indices
Those first two statements match Python functions and methods: min
does the first, and list.index does the second.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SEARCHING 206
We see that list.remove does the third, and the refinement of “find the
index of the new minimum element in the list” is also straightforward:
Download alg/find_remove_find3.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
smallest = min(L)
min1 = L.index(smallest)
L.remove(smallest)
next_smallest = min(L)
min2 = L.index(next_smallest)
return the two indices
Since we removed the smallest element, we need to put it back. Also,
when we remove a value, the indices of the following values shift down
by one. So, when smallest was removed, in order to get the indices of the
two lowest values in the original list, we need to add 1 to min2:
Download alg/find_remove_find4.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
smallest = min(L)
min1 = L.index(smallest)
L.remove(smallest)
next_smallest = min(L)
min2 = L.index(next_smallest)
put smallest back into L
if min1 comes before min2, add 1 to min2
return the two indices
That’s enough refinement (finally!) to do it all in Python:
Download alg/find_remove_find5.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
smallest = min(L)
min1 = L.index(smallest)
L.remove(smallest)
next_smallest = min(L)
min2 = L.index(next_smallest)
L.insert(min1, smallest)
if min1 <= min2:
min2 += 1
return (min1, min2)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SEARCHING 207
That seems like a lot of work, and it is. Even if you go right to code,
you’ll be thinking through all those steps.
Sort, Identify Minimums, Get Indices
We again restate the algorithm, rewritten with one instruction per line:
Download alg/sort_then_find1.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
sort a copy of L
get the two smallest numbers
find their indices in the original list L
return the two indices
That looks straightforward; we can use list.sort, and then the smallest
items will be at the front. Being careful, we notice that we should work
on a copy of L—never change the contents of parameters unless the
docstrings says to!
Download alg/sort_then_find2.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
temp_list = L[:]
temp_list.sort()
smallest = temp_list[0]
next_smallest = temp_list[1]
find their indices in the original list
return the two indices
Now we can find the indices and return them the same way we did in
find-remove-find:
Download alg/sort_then_find3.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
temp_list = L[:]
temp_list.sort()
smallest = temp_list[0]
next_smallest = temp_list[1]
min1 = L.index(smallest)
min2 = L.index(next_smallest)
return (min1, min2)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SEARCHING 208
Walk Through the List
We’ll start the same way as for the first two:
Download alg/walk_through1.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
examine each value in the list in order
keep track of the indices of the two smallest values found so far
update these values when a new smaller value is found
return the two indices
We’ll also move the second line before the first one because it describes
the whole process; it isn’t a single step. Also, when we see phrases like
each value, we think of iteration; the third line is part of that iteration,
so we’ll indent it:
Download alg/walk_through2.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
keep track of the indices of the two smallest values found so far
examine each value in the list in order
update these values when a new smaller value is found
return the two indices
Every loop has three parts: an initialization section to set up the vari-
ables we’ll need, a loop condition, and a loop body. Here, the initializa-
tion will set up min1 and min2, which will be the indices of the smallest
two items; a natural choice is to set them to the first two elements of
the list:
Download alg/walk_through3.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
set min1 and min2 to the indices of the smallest and next-smallest
values at the beginning of L
examine each value in the list in order
update these values when a new smaller value is found
return the two indices
We can turn that first line into a couple lines of code; we’ve left our
English version in as a comment:
Download alg/walk_through4.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SEARCHING 209
# set min1 and min2 to the indices of the smallest and next-smallest
# values at the beginning of L
if L[0] < L[1]:
smallest, next_smallest = 0, 1
else:
smallest, next_smallest = 1, 0
examine each value in the list in order
update these values when a new smaller value is found
return the two indices
We have a couple choices now. We can iterate with a for loop over the
values, a for loop over the indices, or a while loop over the indices. Since
we’re trying to find indices and we want to look at all of the elements in
the list, we’ll use a for loop over the indices—and we’ll start at index 2
because we’ve examined the first two values already. At the same time,
we’ll refine the statement in the body of the loop to mention min1 and
min2.
Download alg/walk_through5.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
# set min1 and min2 to the indices of the smallest and next-smallest
# values at the beginning of L
if L[0] < L[1]:
min1, min2 = 0, 1
else:
min2, min1 = 0, 1
# examine each value in the list in order
for i in range(2, len(values)):
update min1 and/or min2 when a new smaller value is found
return the two indices
Now for the body of the loop. We’ll pick apart “update min1 and/or min2
when a new smaller value is found.”
If L[i] is smaller than both, then we have a new smallest item, so min1
currently holds the second smallest and min2 currently holds the third
smallest. We need to update both of them.
If L[i] is larger than min1 and smaller than min2, we have a new second
smallest.
If L[i] is larger than both, we skip it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SEARCHING 210
Download alg/walk_through6.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
# set min1 and min2 to the indices of the smallest and next-smallest
# values at the beginning of L
if L[0] < L[1]:
min1, min2 = 0, 1
else:
min2, min1 = 1, 0
# examine each value in the list in order
for i in range(2, len(L)):
L[i] is larger than both min1 and min2, smaller than both, or
in between.
if L[i] is larger than both min1 and min2, skip it
if L[i] is smaller than min1 and min2, update them both
if L[i] is in between, update min2
return (min1, min2)
All of those are easily translated to Python; in fact, we don’t even need
code for the “larger than both” case:
Download alg/walk_through7.py
def find_two_smallest(L):
'''Return a tuple of the indices of the two smallest values in list L.'''
# set min1 and min2 to the indices of the smallest and next-smallest
# values at the beginning of L
if L[0] < L[1]:
min1, min2 = 0, 1
else:
min2, min1 = 1, 0
# examine each value in the list in order
for i in range(2, len(L)):
# L[i] is larger than both min1 and min2, smaller than both, or
# in between.
# New smallest?
if L[i] < L[min1]:
min2 = min1
min1 = i
# New second smallest?
elif L[i] < L[min2]:
min2 = i
return (min1, min2)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TIMING 211
10.2 Timing
Profiling a program means measuring how long it takes to run and
how much memory it uses. These two measures—space and time—
are fundamental to the theoretical study of algorithms. They are also
pretty important from a pragmatic point of view. Fast programs are
more useful than slow ones, and programs that need more memory
than what your computer has aren’t particularly useful at all.
We ran the three functions we developed to find the two lowest values
in a list on 1,400 monthly readings of air pressure in Darwin, Australia,
from 1882 to 1998.1 The execution times were as follows:
Algorithm Running Time (ms)
Find, remove, find 1.117
Sort, identify, index 2.128
Walk through the list 1.472
Notice how small these times are. No human being can notice the dif-
ference between one and two milliseconds; if this code never has to
process lists with more than 1,400 values, we would be justified in
choosing an implementation based on simplicity or clarity, rather than
speed.
But what if we wanted to process millions of values? Find-remove-find
outperforms the other two algorithms on 1,400 values, but how much
does that tell us about how they will perform on data sets 1,000 times
larger? That will be the subject of Chapter 11, Searching and Sorting,
on page 214.
10.3 Summary
In this chapter, we learned the following:
• The most effective way to design algorithms is to use top-down
design, in which goals are broken down into subgoals until the
steps are small enough to be translated directly into a program-
ming language.
• The performance of a program can be characterized by how much
time and memory it uses. This can be determined experimentally
by profiling its execution.
1. See http://www.stat.duke.edu/~mw/ts_data_sets.html.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 212
• Almost all problems have more than one correct solution. Choos-
ing between them often involves a trade-off between simplicity and
performance.
• Code that has side effects is harder to understand and debug
than code that does not, so in general functions should return
values rather than modifying their arguments or external data
structures.
10.4 Exercises
Here are some exercises for you to try on your own:
1. A DNA sequence is a string made up of the letters A, T, G, and C.
To find the complement of a DNA sequence, As are replaced by Ts,
Ts by As, Gs by Cs, and Cs by Gs. For example, the complement
of AATTGCCGT is TTAACGGCA.
a) Write an outline in English of the algorithm you would use to
find the complement.
b) Reexamine your algorithm. Will any characters be changed
to their complement and then changed back to their original
value? If so, rewrite your outline. Hint: convert one character
at a time, rather than all of the As, Ts, Gs, or Cs at once.
c) Write the complement(sequence) function.
2. a) Write a loop (including initialization) that uses a loop to find
both the minimum value in a list and that value’s index in
one pass through the list.
b) Write the min_index(sequence) function that returns the mini-
mum value and that value’s index in a tuple.
c) You might also want to find the maximum value and its index.
Write a new function min_or_max_index(sequence, bool) where if
bool is true, it returns the minimum and its index, and where
if bool is false, it returns the maximum and its index.
3. Write a set of nose tests for the find-two-smallest functions. Think
about what kinds of data are interesting, long lists or short lists,
and what order the items are in. Here is one list to test with: [1, 2].
What other interesting ones are there?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 213
4. What happens if the functions to find the two smallest values in a
list are passed a list of length 1? What should happen, and why?
How about length 0? Modify one of the docstrings to describe what
happens.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 11
Searching and SortingA huge part of computer science involves studying how to organize,
store, and retrieve data. Searching and sorting are fundamental parts
of programming, and they are also a good way to explain how computer
scientists compare the efficiency of different algorithms and to explain
what they mean by efficiency. In this chapter, we will develop several
algorithms for searching and sorting lists and then use them to explore
what it means for one algorithm to be faster than another.
11.1 Linear Search
As you have already seen, Python lists have a method called index that
searches for a particular item:
Download searchsort/index_help.txt
index(...)
L.index(value, [start, [stop]]) -> integer -- return first index of value
Method index starts at the front of the list and examines each item in
turn. For reasons that will soon become clear, this technique is called
linear search. We’ll examine three different but related versions.
Basic Linear Search
Here’s the first version of linear search; this is a common version:
Download searchsort/linear_search_1.py
def linear_search(v, L):
'''Return the index of the first occurrence of v in list L, or return len(L)
if v is not in L.'''
i = 0
Prepared exclusively for Trieu Nguyen
LINEAR SEARCH 215
# Keep going until we reach the end of L or until we find v.
while i != len(L) and L[i] != v:
i = i + 1
return i
This version uses variable i as the current index and marches through
the values in L. The first check in the loop condition, i != len(L), makes
sure we look at only valid indices; if we were to omit that check, then if
v is not in L, we would end up trying to access L[len(L)]. This, of course,
would not be valid.
The second check, L[i] != v, causes the loop to exit when we find v. The
loop body increments i; we enter the loop when we haven’t reach the
end of L and when L[i] isn’t the value we are looking for.
At the end, we return i’s value, which is either the index of v (if the
second loop check was false) or is len(L) if v was not in L.
for loop Version of Linear Search
The first version requires two checks each time through the loop. The
first check (i != len(L)) is almost unnecessary; it evaluates to true almost
every time through the loop, so the only effect it has is to make sure we
don’t attempt to index past the end of the list.
Here’s the second version, where we try to get around this problem
using Python’s for loop:
Download searchsort/linear_search_2.py
def linear_search(v, L):
'''Return the index of the first occurrence of v in list L, or return len(L)
if v is not in L.'''
i = 0
for value in L:
if value == v:
return i
i += 1
return len(L)
With this version, we no longer need the first check because the for loop
controls the number of iterations. This for loop version is significantly
faster than basic linear search; we’ll see in a bit how much faster.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LINEAR SEARCH 216
Sentinel Search
The last linear search we will study is called sentinel search.1 Remem-
ber that a problem with basic linear search is that we check i != len(L)
every time through the loop even though it can never be false except
when v is not in L. So, we’ll play a trick: we’ll add v to the end of L before
searching so that we’re guaranteed to find it! And we need to remove it
before the function exits so that the list looks unchanged at the end:
Download searchsort/linear_search_3.py
def linear_search(v, L):
'''Return the index of the first occurrence of v in list L, or return len(L)
if v is not in L.'''
# Add the sentinel.
L.append(v)
i = 0
# Keep going until we find v.
while L[i] != v:
i = i + 1
# Remove the sentinel.
L.pop()
return i
Timing the Searches
Here is a program that we used to time the three searches on a list with
1 million values:
Download searchsort/linear_time_1.py
import time
import linear_search_1
import linear_search_2
import linear_search_3
def time_it(search, v, L):
'''Time how long it takes to run function search to find value v in list L.'''
t1 = time.time()
search(v, L)
t2 = time.time()
return (t2 - t1) * 1000.
1. A sentinel is a guard whose job it is to stand watch.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
LINEAR SEARCH 217
def print_times(v, L):
'''Print the number of milliseconds it takes for linear_search(v, L)
to run for list.index, basic linear search, the for loop linear search,
and sentinel search.'''
# Get list.index's running time.
t1 = time.time()
L.index(v)
t2 = time.time()
index_time = (t2 - t1) * 1000.
# Get the other three running times.
basic_time = time_it(linear_search_1.linear_search, v, L)
for_time = time_it(linear_search_2.linear_search, v, L)
sentinel_time = time_it(linear_search_3.linear_search, v, L)
print "%d\t%.02f\t%.02f\t%.02f\t%.02f" % \
(v, basic_time, for_time, sentinel_time, index_time)
L = range(1000001)
linear_search_1.linear_search(10, L)
print_times(10, L)
print_times(500000, L)
print_times(1000000, L)
Function time_it runs whatever search function it is given on v and L
and returns how long the search takes. Function print_times calls time_it
with the various linear search functions we have been exploring and
prints the search times.
Linear Search Running Time
The running times of the three linear searches with that of Python’s
list.index are compared in Figure 11.1, on the following page. This com-
parison used a list of 1,000,001 items and three test cases: an item
near the front, an item roughly in the middle, and the last item. Except
for the first case, where the speeds differ by only a few percent points,
our basic linear search takes about seven times as long as the one built
into Python, and the for loop search and sentinel search take about four
times as long.
What is more interesting is the way the running times of these func-
tions increase with the number of items they have to examine. Roughly
speaking, when they have to look through twice as much data, every
one of them takes twice as long. This is reasonable, because indexing
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BINARY SEARCH 218
Case basic for sentinel list.index
First 0.01 0.01 0.03 0.01
Middle 138 69 62 17
Last 273 139 124 35
Figure 11.1: Running times for linear search, milliseconds
a list, adding 1 to an integer, and evaluating the loop control expres-
sion require the computer to do a fixed amount of work. Doubling the
number of times the loop has to be executed therefore doubles the total
number of operations, which in turn should double the total running
time. This is why this kind of search is called linear: the time to do it
grows linearly with the amount of data being processed.
11.2 Binary Search
Is there a faster way to find values than linear search? The answer is
yes—we can do much better, provided the list is sorted.
To understand how, think about finding a name in a phone book. You
open the book in the middle, glance at a name on the page, and imme-
diately know which half to look in next. After checking only two names,
you have eliminated 34 of the numbers in the phone book. Even in a
large city like Toronto, whose phone book has hundreds of thousands
of entries, finding the name you want takes only a few steps.
This technique is called binary search, because each step divides the
remaining data into two equal parts: values that come before the one
being looked for and values that come after it. To figure out how fast it
is, think about how big a list can be searched in a fixed number of steps.
One step divides two values; two steps divide four; three steps divide 23
= 8, four divide 24 = 16, and so on. Turning this around, N values can
be searched in roughly log2N steps.2 As shown in Figure 11.2, on the
next page, this increases much less quickly than the time needed for
linear search.
2. More exactly, N values can be searched in ⌈log2N⌉ steps, where ⌈⌉ is the ceiling func-
tion that rounds a value up to the nearest integer.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BINARY SEARCH 219
N Steps Required
100 7
1000 10
1,0000 14
10,0000 17
100,000 24
1,000,000 27
Figure 11.2: Logarithmic growth
The key to binary search is to keep track of three parts of the list: the left
part, which contains values that come before the value we are searching
for; the right part, which contains values that come after the value we
are searching for; and the middle part, which contains values that we
haven’t yet examined—the unknown section. We’ll use two variables to
keep track of the boundaries: i will mark the index of the first unknown
value, and j will mark the index of the last unknown value.
At the beginning of the algorithm, the unknown section makes up the
entire list, so we will set i to 0 and j to the length of the list minus
one. We are done when that unknown section is empty—when we’ve
examined every item in the list. This happens when i == j + 1—when the
values cross. (When i == j, there is still one item left in the unknown
section.)
To make progress, we will set either i or j to near the middle of the range
between them, which is at (i + j) / 2.
Think for a moment about the value at (i + j) / 2. If it is less than v, we
need to move i up, while if it is greater than j, we should move j down.
But where exactly do we move them? When we move i up, we don’t want
to set it to the midpoint exactly, because L[i] is not included in the range;
instead, we set it to one past the middle, in other words, to (i + j) / 2 + 1.
Similarly, when we move j down, we move it to (i + j) / 2 - 1.
The completed function is as follows:
Download searchsort/binary_search.py
def binary_search(v, L):
"""Return the index of the leftmost occurrence of v in list L, or -1 if
v is not in L."""
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BINARY SEARCH 220
# Mark the left and right indices of the unknown section.
i = 0
j = len(L) - 1
while i != j + 1:
m = (i + j) / 2
if L[m] < v:
i = m + 1
else:
j = m - 1
if 0 <= i < len(L) and L[i] == v:
return i
else:
return -1
Of course, no code is really complete until it has been tested. We tested
the various linear searches but did not show the tests; they are similar
to these:
Download searchsort/binary_test.py
'''Test binary search.'''
import nose
from binary_search import binary_search
# The list to search with.
VALUES = [1, 3, 4, 4, 5, 7, 9, 10]
def test_first():
'''Test a value at the beginning of the list.'''
assert binary_search(1, VALUES) == 0
def test_duplicate():
'''Test a duplicate value.'''
assert binary_search(4, VALUES) == 2
def test_middle():
'''Test searching for the middle value.'''
assert binary_search(5, VALUES) == 4
def test_last():
'''Test searching for the last value.'''
assert binary_search(10, VALUES) == 7
def test_missing_start():
'''Test searching for a missing value at the start.'''
assert binary_search(-3, VALUES) == -1
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BINARY SEARCH 221
Case list.index binary_search Ratio
First 0.03 0.05 0.66
Middle 107 0.04 2643
Last 15.73 0.04 (Wow!) 4304
Figure 11.3: Running times for binary search
def test_missing_middle():
'''Test searching for a missing value in the middle.'''
assert binary_search(2, VALUES) == -1
def test_missing_end():
'''Test searching for a missing value at the end.'''
assert binary_search(11, VALUES) == -1
if __name__ == '__main__':
nose.runmodule()
Binary Search Running Time
Binary search is more complicated to write and understand than linear
search. Is it fast enough to make the extra effort worthwhile? To find
out, we can compare it to list.index. As before, we search for the first,
middle, and last items in a list with 100,001 elements.
The results are impressive (see Figure 11.3). Binary search is up to sev-
eral thousand times faster than its linear counterpart when searching 1
million items. Most important, if we double the number of items, binary
search takes only one more iteration, while the time for list.index nearly
doubles. Note also that although the time taken for linear search grows
in step with the index of the item found, there is no such pattern for
binary search. No matter where the item is, it takes the same number
of steps.
Built-in Binary Search
The Python standard library’s bisect module includes binary search
functions that are slightly faster than our binary search. bisect_left re-
turns the index where an item should be inserted in a list to keep it
in sorted order, assuming it is sorted to begin with. insort_left actually
does the insertion. The word left in their name signals that they find
the leftmost (lowest index) position where they can do their jobs; the
complementary functions bisect_right and insort_right find the rightmost.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SORTING 222
563 7590 1708 2142 3323 6197 1985 1316 1824 472
1346 6029 2670 2094 2464 1009 1475 856 3027 4271
3126 1115 2691 4253 1838 828 2403 742 1017 613
3185 2599 2227 896 975 1358 264 1375 2016 452
3292 538 1471 9313 864 470 2993 521 1144 2212
2212 2331 2616 2445 1927 808 1963 898 2764 2073
500 1740 8592 10856 2818 2284 1419 1328 1329 1479
Figure 11.4: Acres lost to forest fires in Canada (in 000s),
1918–87
11.3 Sorting
Now let’s look at a slightly harder problem. Figure 11.4, taken from
[Hynnd], shows the number of acres burned in forest fires in Canada
from 1918 to 1987. One way to find out how much forest was destroyed
in the N worst years is to sort the list and then take the last N values
(Section 10.1, Sort, Identify Minimums, Get Indices, on page 207):
Download searchsort/sort1.py
def find_largest(N, L):
"""Return the N largest values in L in order from smallest to largest."""
copy = L[:]
copy.sort()
return copy[:N]
This algorithm is short, clean, and easy to understand, but it relies on
a bit of black magic. How does list.sort work, anyway? And how efficient
is it?
It turns out that many sorting algorithms have been developed over the
years, each with its own strengths and weaknesses. Broadly speaking,
they can be divided into two categories: those that are simple but inef-
ficient and those that are efficient but harder to understand and imple-
ment. We’ll examine two of the former kind; the rest rely on techniques
typically taught in a second course.
Selection Sort
Selection sort works by partitioning the list in two. The section at the
front contains values that are now in sorted order; the one at the back
contains values that have yet to be sorted. As you can probably guess
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SORTING 223� �2 3
'a' 'b' 'c' 'd'
L
4 5
'k' 'g'
sorted
6
'l'
7
'm'
8
'e'
9
'h'
10
'f'
11
'j'
12
'i'
unsorted
next smallest
0 1 2 3
'a' 'b' 'c' 'd'
L
4 5
'e' 'g'
sorted
6
'l'
7
'm'
8
'k'
9
'h'
10
'f'
11
'j'
12
'i'
unsorted
Before
After
next smallest
Figure 11.5: Selection sort
from this description, selection sort works by repeatedly finding the
next smallest item in the unsorted section and placing it at the end of
the sorted section (see Figure 11.5). This works because we are select-
ing the items in order. We first select the smallest item and move it to
the front; then we select the second-smallest item and move it to the
second spot, and so on.
Here’s the selection sort algorithm, partially in English:
Download searchsort/sort2.py
def selection_sort(L):
"""Reorder the items in L from smallest to largest."""
i = 0
while i != len(L):
# Find the index of the smallest item in L[i:].
# Swap that smallest item with L[i].
i = i + 1
In this algorithm, i acts as a marker for the beginning of the unsorted
section.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SORTING 224
We can replace the second comment with a single line of code:3
Download searchsort/sort3.py
def selection_sort(L):
"""Reorder the values in L from smallest to largest."""
i = 0
while i != len(L):
# Find the index of the smallest item in L[i:].
L[i], L[smallest] = L[smallest], L[i]
i = i + 1
Now all that’s left is finding the index of the smallest item in L[i:]. This
is complex enough that it’s worth putting it in a function of its own:
Download searchsort/sort4.py
def selection_sort(L):
"""Reorder the values in L from smallest to largest."""
i = 0
while i != len(L):
smallest = find_min(L, i)
L[i], L[smallest] = L[smallest], L[i]
i = i + 1
def find_min(L, b):
"""Return the index of the smallest value in L[b:]."""
smallest = b # The index of the smallest so far.
i = b + 1
while i != len(L):
if L[i] < L[smallest]:
# We found a smaller item at L[i].
smallest = i
i = i + 1
return smallest
Function find_min runs along L[b:], keeping track of the index of the min-
imum item so far in variable smallest. When it finds a smaller item, it
updates smallest.
3. Most Python programmers would probably write the loop header as for i in range(len(L)),
rather than incrementing i explicitly in the body of the loop. We’re doing the latter here to
make it clearer when the loop ends and to show why.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SORTING 225
As always, our function isn’t done until it has been tested:
Download searchsort/sort4test.py
from sort4 import selection_sort
import nose
def run_test(original, expected):
'''Sort list original and compare it to list expected.'''
selection_sort(original)
assert original == expected
def test_empty():
'''Test sorting empty list.'''
run_test([], [])
def test_one():
'''Test sorting a list of one value.'''
run_test([1], [1])
def test_two_ordered():
'''Test sorting an already-sorted list of two values.'''
run_test([1, 2], [1, 2])
def test_two_reversed():
'''Test sorting a reverse-ordered list of two values.'''
run_test([2, 1], [1, 2])
def test_three_identical():
'''Test sorting a list of three equal values.'''
run_test([3, 3, 3], [3, 3, 3])
def test_three_split():
'''Test sorting a list with an odd value out.'''
run_test([3, 0, 3], [0, 3, 3])
if __name__ == '__main__':
nose.runmodule()
Insertion Sort
Like selection sort, insertion sort keeps a sorted section at the beginning
of the list. Rather than scan all of the unsorted section for the next
smallest item, though, it takes the next item from the unsorted section
and inserts it where it belongs in the sorted section.
In outline, this is as follows:
Download searchsort/sort5.py
def insertion_sort(L):
"""Reorder the values in L from smallest to largest."""
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SORTING 226
i = 0
while i != len(L):
# Insert L[i] where it belongs in L[0:i+1].
i = i + 1
This is exactly the same code as for selection sort; the difference is in
the comment in the loop.
How does insert work? It works by finding out where L[b] belongs and
then moving it. Where does it belong? It belongs after every value less
than or equal to it and before every value that is greater than it.4 We
need the check i != 0 in case L[b] is smaller than every value in L[0:b]:
Download searchsort/sort7.py
def insertion_sort(L):
"""Reorder the values in L from smallest to largest."""
i = 0
while i != len(L):
insert(L, i)
i = i + 1
def insert(L, b):
"""Insert L[b] where it belongs in L[0:b + 1];
L[0:b - 1] must already be sorted."""
# Find where to insert L[b] by searching backwards from L[b] for a smaller item.
i = b
while i != 0 and L[i - 1] >= L[b]:
i = i - 1
# Move L[b] to index i, shifting the following values to the right.
value = L[b]
del L[b]
L.insert(i, value)
This passes all the tests we wrote earlier for selection sort.
Performance
We have two algorithms of roughly equal complexity. Which should we
use?
4. We could equally well say “After every value less than it and before every value that is
greater than or equal to it.” This would produce a slightly different function, but the end
result would still be a sorted list.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SORTING 227
It’s easy enough to write a program to compare their running times,
along with that for list.sort:
Download searchsort/sort_time.py
import time
from sort4 import selection_sort
from sort7 import insertion_sort
from ms import mergesort
def built_in(L):
'''Call list.sort --- we need our own function to do this
so that we can treat it as we treat our own sorts.'''
L.sort()
def print_times(L):
'''Print the number of milliseconds it takes for selection sort
and insertion sort to run.'''
print len(L),
for func in (selection_sort, insertion_sort, mergesort, built_in):
if func in (selection_sort, insertion_sort) and len(L) > 4000:
continue
L_copy = L[:]
t1 = time.time()
func(L_copy)
t2 = time.time()
print "\t%.1f" % ((t2 - t1) * 1000.),
for list_size in [10, 1000, 2000, 3000, 4000, 5000, 10000, 20000, 40000, 80000]:
L = range(list_size)
L.reverse()
print_times(L)
The results are shown in Figure 11.6, on the following page. Something
is very clearly wrong, because our sorting functions are tens of thou-
sands of times slower than the built-in function. What’s more, the time
required by our routines is growing faster than the size of the data.
On 1,000 items, for example, selection sort takes about 0.5 millisec-
onds per item, but on 10,000 items, it needs about 4 milliseconds per
item—an eightfold increase! What is going on?
To answer this, we examine what happens in the inner loops of our
two algorithms. On the first iteration of selection sort, the inner loop
examines every element to find the smallest. On the second iteration, it
looks at all but one; on the third, it looks at all but two, and so on.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MORE EFFICIENT SORTING ALGORITHMS 228
List Length Selection Sort Insertion Sort list.sort
10 0.1 0.1 0.0
1000 481 300 0.1
2000 1640 1223 0.2
3000 3612 2772 0.4
4000 6536 4957 0.6
5000 10112 7736 0.6
10000 40763 32382 1.3
Figure 11.6: Running times for selection and insertion sort,
milliseconds
If there are N items in the list, then the number of iterations of the
inner loop, in total, is roughly N + (N-1) + (N-2) + ... + 1, or N(N+1)2 .
Putting it another way, the number of steps required to sort N items is
roughly proportional to N2+N. For large values of N, we can ignore the
second term and say that the time needed by selection sort grows as
the square of the number of values being sorted. And indeed, examining
the timing data further shows that doubling the size of the list increases
the running time by four.
The same analysis can be used for insertion sort, since it also examines
one element on the first iteration, two on the second, and so on. (It’s just
examining the already-sorted values, rather than the unsorted values.)
So, why is insertion sort slightly faster? The reason is that, on average,
only half of the values need to be scanned in order to find the location
in which to insert the new value, while with selection sort, every value
in the unsorted section needs to be examined in order to select the
smallest one. But, wow, list.sort is so much faster!
11.4 More Efficient Sorting Algorithms
The analysis of selection and insertion sort begs the question: how can
list.sort be so much more efficient? The answer is the same as it was
for binary search: by taking advantage of the fact that some values are
already sorted.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MERGESORT: AN N log2N ALGORITHM 229
N N2 N log2N
10 100 3.32
100 10,000 6.64
1000 1,000,000 9.96
Figure 11.7: Sorting times
A First Attempt
Consider the following function:
Download searchsort/binsort.py
import bisect
def bin_sort(values):
'''Sort values, creating a new list.'''
result = []
for v in values:
bisect.insort_left(result, v)
return result
This code uses bisect.insort_left to figure out where to put each value
from the original list into a new list that is kept in sorted order. As we
have already seen, doing this takes time proportional to log2N, where N
is the length of the list. Since N values have to be inserted, the overall
running time ought to be N log2N. As shown in Figure 11.7, this grows
much more slowly with the length of the list than N2.
Unfortunately, there’s a flaw in this analysis. It’s correct to say that
bisect.insort_left needs only log2N time to figure out where to insert a
value, but actually inserting it takes time as well. To create an empty
slot in the list, we have to move all the values above that slot up one
place. On average, this means copying half of the list’s values, so the
cost of insertion is proportional to N. Since there are N values to insert,
our total time is N(N + log2N). For large values of N, this is once again
roughly proportional to N2.
11.5 Mergesort: An Nlog2N Algorithm
There are several well-known fast sorting algorithms; mergesort, quick-
sort, and heapsort are the ones you are most likely to encounter in a
future CS course. Most of them involve techniques that we haven’t yet
taught you, but mergesort can be written to be more accessible. Merge-
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MERGESORT: AN N log2N ALGORITHM 230
sort is built around the idea that taking two sorted lists and merging
them is proportional to the number of items in both lists. We’ll start
with very small lists and keep merging them until we have a single
sorted list.
Merging Two Sorted Lists
Given two lists L1 and L2, we can produce a new sorted list by running
along L1 and L2 and comparing pairs of elements.
Here is the code for merge:
Download searchsort/merge.py
def merge(L1, L2):
"""Merge sorted lists L1 and L2 and return the result."""
newL = []
i1 = 0
i2 = 0
# For each pair of items L1[1], L2[i2], copy the smaller into newL.
while i1 != len(L1) and i2 != len(L2):
if L1[i1] <= L2[i2]:
newL.append(L1[i1])
i1 += 1
else:
newL.append(L2[i2])
i2 += 1
# Gather any leftover items from the two sections.
# Note that one of them will be empty because of the loop condition.
newL.extend(L1[i1:])
newL.extend(L2[i2:])
return newL
i1 and i2 are the indices into L1 and L2, respectively; in each iteration,
we compare L1[i1] to L2[i2] and copy the smaller item to the resulting list.
At the end of the loop, we have run out of items in one of the two lists,
and the two extend calls will append the rest of the items to the result.
Mergesort
Here is the header for mergesort:
Download searchsort/mergesort_header.py
def mergesort(L):
"""Sort L in increasing order."""
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MERGESORT: AN N log2N ALGORITHM 231
0 1 2 3
workspace
4 � �i
Figure 11.8: List of lists in mergesort
Mergesort uses function merge to do the bulk of the work. Here is the
algorithm, which creates and keeps track of a list of lists:
• Take list L, and make a list of one-item lists from it.
• As long as there are two lists left to merge, merge them, and
append the new list to the list of lists.
The first step is straightforward:
Download searchsort/mergesort_make_list.py
# Make a list of 1-item lists so that we can start merging.
workspace = []
for i in range(len(L)):
workspace.append([L[i]])
The second step is trickier. If we remove the two lists, then we’ll run
into the same problem that we ran into in binsort: all the following lists
will need to shift over, which takes time proportional to the number of
lists.
Instead, we’ll keep track of the index of the next two lists to merge.
Initially, they will be at indices 0 and 1, and then 2 and 3, and so on
(Figure 11.9, on page 233). Here’s our refined algorithm:
• Take list L, and make a list of one-item lists from it.
• Start index i off at 0.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MERGESORT: AN N log2N ALGORITHM 232
• As long as there are two lists, at indices i and i + 1, merge them,
append the new list to the list of lists, and increment i by 2.
With that, we can go straight to code:
Download searchsort/mergesort_function.py
def mergesort(L):
"""Sort L."""
# Make a list of 1-item lists so that we can start merging.
workspace = []
for i in range(len(L)):
workspace.append([L[i]])
# The next two lists to merge are workspace[i] and workspace[i + 1].
i = 0
# As long as there are at least two more lists to merge, merge them.
while i < len(workspace) - 1:
L1 = workspace[i]
L2 = workspace[i + 1]
newL = merge(L1, L2)
workspace.append(newL)
i += 2
# Copy the result back into L.
if len(workspace) != 0:
L[:] = workspace[-1][:]
Notice that, since we’re always making new lists, we need to copy the
last of the merged lists back into parameter L.
Mergesort Analysis
Mergesort, it turns out, is N log2N, where N is the number of items in L.
It may help to refer to Figure 11.9, on the next page.
The first part of the function, creating the list of one-item lists, takes N
iterations, one for each item.
The second loop, in which we continually merge lists, will take some
care to analyze. We’ll start with the very last iteration, in which we are
merging two lists with about N
2 items. As we’ve seen, function merge
copies each element into its result exactly once, so with these two lists,
this merge step takes roughly N steps.
On the previous iteration, there are two lists of size N
4 to merge into
one of the two lists of size N
2 , and on the iteration before that there are
another two lists of size N
2 to merge into the second list of size N
2 . Each
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 233
Merge
Merge Merge
Merge Merge Merge Merge
Figure 11.9: Steps in mergesort
of these two merges takes roughly N
2 steps, so the two together take
roughly N steps in total.
On the iteration before that, there are a total of eight lists of size N
8 to
merge into the four lists of size N
4 . Four merges of this size together also
take roughly N steps.
We can subdivide a list with N items a total of log2N times, using an
analysis much like we used for binary search. Since at each “level”
there are a total of N items to be merged, each of these log2N levels takes
roughly N steps. Hence, mergesort take time proportional to N log2N.
11.6 Summary
In this chapter, we learned the following:
• Linear search is the simplest way to find a value in a list, but on
average, the time required is directly proportional to the length of
the list.
• Binary search is much faster—the average time is proportional to
the logarithm of the list’s length—but it works only if the list is in
sorted order.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 234
Big-Oh and All That
Our method of analyzing the performance of searching andsorting algorithms might seem like hand-waving, but there isactually a well-developed mathematical theory behind it. If fand g are functions, then the expression f(x) = O(g(x)) is read“f is big-oh of g” and means that for sufficiently large valuesof x, f(x) is bounded above by some constant multiple of g(x),or equivalently that the function g gives us an upper boundon the values of the function f. Computer scientists use this togroup algorithms into families, such as those sorting functionsthat execute in N2 time and those that execute in Nlog2N time.
These distinctions have important practical applications. In par-ticular, one of the biggest puzzles in theoretical computer sci-ence today is whether two families of algorithms (called P andNP for reasons that we won’t go into here) are the same or not.Almost everyone thinks they are not, but no one has been ableto prove it (despite the offer of a million-dollar prize for the firstcorrect proof). If it turns out that they are the same, then manyof the algorithms used to encrypt data in banking and militaryapplications (as well as on the Web) will be much more vulner-able to attack than expected.
• Similarly, the average running time of simple sorting algorithms
like selection sort is proportional to the square of the input size N,
while the running time of more complex sorting algorithms grows
as N log2N.
• Looking at how the running time of an algorithm grows as a func-
tion of the size of its inputs is the standard way to analyze and
compare algorithms’ efficiency.
11.7 Exercises
Here are some exercises for you to try on your own:
1. Binary search is significantly faster than the built-in search but
requires that the list is sorted. As you know, the running time for
the best sorting algorithm is on the order of N log2N, where N is the
length of the list. If we search a lot of times on the same list of data,
it makes sense to sort it once before doing the searching; roughly
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 235
how many times do we need to search in order to make sorting
and then searching faster, instead of using the built-in search?
2. Given the unsorted list [6, 5, 4, 3, 7, 1, 2], show what the contents
of the list would be after each iteration of the loop as it is sorted
using the following:
a) Selection sort
b) Insertion sort
3. Another sorting algorithm is bubble sort. Bubble sort involves
keeping a sorted section at the end of the list. The list is tra-
versed, pairs of elements are compared, and larger elements are
swapped into the higher position. This is repeated until all element
are sorted.
a) Using the English description of bubble sort, write an outline
of the bubble sort algorithm in English.
b) Continue using top-down design until you have a Python algo-
rithm.
c) Turn it into a function bubble_sort(L).
d) Write Nose test cases for bubble_sort.
4. In the description of bubble sort in the previous question, the
sorted section of the list was at the end of the list. In this question,
bubble sort will maintain the sorted section of the beginning of the
list. Make sure that you are still implementing bubble sort!
a) Rewrite the English description of bubble sort from the pre-
vious question with the necessary changes so that the sorted
elements are at the beginning of the list instead of the end.
b) Using your English description of bubble sort, write an outline
of the bubble sort algorithm in English.
c) Write the bubble_sort_2(L) function.
d) Write Nose test cases for bubble_sort_2.
5. Modify the timing program to compare bubble sort with insertion
and selection sort. Explain the results.
6. Modify the timing program to compare Python’s built-in list.sort
method with insertion and selection sort. Clearly, the built-in sort
does not use insertion, selection, or bubble sort; instead, some
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 236
advanced tricks are being used in order to get a huge speedup.
You will learn these tricks in your next computer science course.
7. The analysis of bin_sort said, “Since N values have to be inserted,
the overall running time is N log2N.” Point out a flaw in this rea-
soning, and explain whether it affects the overall conclusion.
8. There are at least two ways to come up with loop conditions. One of
them is to answer the question “When is the work done?” and then
negate it. In function merge in Section 11.5, Merging Two Sorted
Lists, on page 230, the answer is “when we run out of items in one
of the two lists,” which is described by this expression: i1 == len(L1)
or i2 == len(L2). Negating this leads to our condition i1 != len(L1) and
i2 != len(L2).
Another way to come up with a loop condition is to ask “What are
the valid values of the loop index?” In function merge, the answer
to this is 0 <= i1 < len(L1) and 0 <= i2 < len(L2); since i1 and i2 start a
zero, we can drop the comparisons with zero, giving us i1 < len(L1)
and i2 < len(L2).
Is there another way to do it? Have you tried both approaches?
Which do you prefer?
9. In function mergesort in Section 11.5, Mergesort, on page 230,
there are two calls to extend. They are there because, when the
preceding loop ends, one of the two lists still has items in it that
have not been processed. Rewrite that loop so that these extend
calls are not needed.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 12
ConstructionA good scientist can do a lot with a handheld magnifying glass and a
pair of tweezers. However, she can do a lot more with a fully equipped
lab. The same is true for programming. In theory, you can write any pro-
gram at all using nothing but arithmetic and conditionals, but useful
software is a lot easier to build if you have more tools at your disposal.
This chapter will introduce a few of those tools. Some are part of the
Python language and will allow you to code something using a couple
of statements that might otherwise take up an entire page. Others are
mental tools, such as testing and debugging techniques, ways to handle
errors, and patterns worth naming because they come up so frequently.
Together, they help you become more fluent as a programmer.
12.1 More on Functions
Good programmers break their code down into functions that can be
reused. Therefore, it’s not surprising that Python (as well as other lan-
guages) provides a number of tools for making functions more flexible.
Default Parameter Values
Default parameter values are the first of these tools. Here’s an example:
Download construct/default_define.py
def total(values, start=0, end=None):
'''Add up the values in a list. If none are given, the total is zero. If
'start' is not specified, start at the beginning. If 'end' is specified,
go up to but not including that index; otherwise, go to the end of the
list.'''
Prepared exclusively for Trieu Nguyen
MORE ON FUNCTIONS 238
if not values:
return 0
if end is None:
end = len(values)
result = 0
for i in range(start, end):
result += values[i]
return result
This function has three parameters: a list of values and two indices that
specify which part of the list to sum up. What makes it different is that
start and end are “assigned” default values in the definition. If the caller
doesn’t override them, Python uses those defaults when the function
is called. This means that the function can be called with one, two, or
three parameters, as shown here:
Download construct/default_call.cmd
>>> numbers = [10, 20, 30]
>>> print "total(numbers, 0, 3):", total(numbers, 0, 3)
total(numbers, 0, 3): 60
>>> print "total(numbers, 2):", total(numbers, 2)
total(numbers, 2): 30
>>> print "total(numbers):", total(numbers)
total(numbers): 60
Specifying default parameter values often makes functions easier to
use. If we didn’t let callers specify start and end indices for total, they
would have to slice their lists whenever they wanted to add up only a
subset of values:
most = total(values[1:-1]) # add up all but the first and last
This is no extra typing but is less efficient, since slicing copies the
values in the list. On the other hand, if total always required callers to
specify start and end indices, then the most common case—summing
up the whole list—would require more typing:
all = total(values, 0, len(values)) # more typing and harder to read
When you give a function’s parameters default values, parameters with-
out default values must come before those that that do. This is because
when a function is called, values are matched with parameters from left
to right; defaults are used only for those that are “left over” at the end of
the parameter list. If Python were to let you define a function like this:
# Illegal!
def total(start=0, values, end=None):
...body as shown before...
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MORE ON FUNCTIONS 239
then the call total(x) could be interpreted as either of these:
1. Assign the value of x to start, and then report an error because
nothing has been assigned to values.
2. Assign 0 to start and the value of x to values.
The second is probably what the caller wanted, but the computer can’t
know that. Worse, with more parameters, there is a greater chance of
confusing one parameter (or one kind of parameter) for another. Thus,
the left-to-right rule described earlier lets everyone involved—the func-
tion’s creator and its user—keep track of which values will be assigned
to which parameters.
One note before we leave this example: if you look closely at total’s def-
inition, you’ll see that start’s default value is 0 (which is a legal index),
but end’s is None (which is not). Why not assign end a more use-
ful default, such as len(values)? The answer is that we can’t, because
Python evaluates the default value when the function is defined, and at
that point, we don’t know what list is going to be passed in for values.
We will discuss the implications of this in more detail in the exercises.
Variable Parameter Lists
Default parameter values are useful, but they still aren’t enough to let
us write functions like max, which can take any number of arguments
at all. We could do this by requiring callers to pass a list of values, but
Python can actually do this for us. Take a close look at the parameter
in this function:
Download construct/our_max.cmd
>>> def our_max(*values):
... '''Find the maximum of any number of values.'''
...
... if not values:
... return None
... m = values[0]
... for v in values[1:]:
... if v > m:
... m = v
... return m
...
>>> our_max(1)
1
>>> our_max(1, 2)
2
>>> our_max(3, 1, 2, 5, 4, -17)
5
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MORE ON FUNCTIONS 240
3
values
0
1
1
2
2
5
3
4
4
-17
5
our_max(3,1,2,5,4,-17)
1
values
0
2
1
our_max(1,2)
values
our_max()
Figure 12.1: How variable arguments work
The key here is the * in front of values. This tells Python to take all the
arguments passed into the call, put them in a tuple, and assign that
tuple to values. For example, if we call our_max(1, 2), values is assigned
(1, 2); when our_max is called with six values, values is assigned a six-
element tuple, while if no arguments are provided, values is assigned an
empty tuple (see Figure 12.1).
A starred parameter does not always have to appear on its own.
Functions are free to define any number of regular parameters before
it. These are matched up with the actual arguments as usual, and
anything left over is then placed in the tuple assigned to the starred
parameter:
Download construct/append_all.cmd
>>> def append_all(old, *new):
... for n in new:
... old.append(n)
... return old
...
>>> values = []
>>> append_all(values, 1, 2, 3)
[1, 2, 3]
>>> append_all(values) # not actually appending anything
[1, 2, 3]
One thing Python doesn’t allow is two or more starred arguments in a
single function, because matching values to them would be ambiguous.
For example, if a function is defined as follows:
# Illegal!
def f(*front, *back):
...body of function...
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MORE ON FUNCTIONS 241
and then called with f(1, 2, 3), there are four possible assignments of the
actual arguments to the parameters:
• a=(1, 2, 3), b=()
• a=(1, 2), b=(3,)
• a=(1,), b=(2, 3)
• a=(), b=(1, 2, 3)
Python also doesn’t allow “regular” parameters to be defined after a
starred parameter:
# Illegal!
def f(*front, y, z):
...body of function...
There’s no technical reason to disallow this, because Python could
match actual arguments to parameters right-to-left instead of left-to-
right and then put the remainder in the starred parameter. On the
other hand, there’s no particular reason to allow it, either. Everything
that can be done this way can be done by defining the starred parame-
ter at the end of the parameter list.
Named Parameter
Python provides yet another way to call functions. Instead of passing
arguments in a particular order, they can be paired with parameters by
name, like this:
Download construct/named_params.cmd
>>> def describe_creature(name, species, age, weight):
... return '%s (%s): %d years, %d kg' % (name, species, age, weight)
...
>>> print describe_creature(name='Charles Darwin', species='Homo sapiens',
age=28, weight=70)
Charles Darwin (Homo sapiens): 28 years, 70 kg
Why would we do this? Because if we specify arguments by name, we
can put them in any order we want and get the same result:
Download construct/named_params_2.cmd
>>> print describe_creature(weight=70, species='Homo sapiens',
age=28, name='Charles Darwin')
Charles Darwin (Homo sapiens): 28 years, 70 kg
Being able to do this is particularly useful when working with functions
that have a very large number of parameters. As we will see in Chap-
ter 14, Graphical User Interfaces, on page 294, for example, a function
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 242
that displays text in a GUI can take a dozen or more parameters to spec-
ify font, weight, colors, borders, and so on. In these cases, it is easier
(and more readable) to specify the parameter names explicitly than to
remember whether the border thickness comes before the border color,
or vice versa.
12.2 Exceptions
Our next tool is a standard way of dealing with errors in programs.
As you have no doubt discovered by now, Python doesn’t like it when
you try to use an out-of-bounds index into a list, use a key not in a
dictionary, or use a string where you need an integer.
In each of these cases, Python reports the error by creating an exception
that contains information about what went wrong. The program can
deal with the exception right where it occurs or handle it somewhere
else; it can also handle exceptions occurring in many different places in
one location. As we’ll see, this allows programmers to separate “normal”
code from error-handling code, which makes both easier to read.
try and except
Consider this short program:
Download construct/simpletry.cmd
>>> try:
... x = 1/0.3
... print 'reciprocal of 0.3 is', x
... x = 1/0.0
... print 'reciprocal of 0.0 is', x
... except:
... print 'error: no reciprocal'
reciprocal of 0.3 is 3.33333333333
error: no reciprocal
The keywords try and except are used for error handling, just like if and
else are used for conditionals. If nothing goes wrong, Python executes
the code in the try block and then skips over the except block entirely
(see Figure 12.2, on the following page). If any exceptions are raised in
the try block, Python immediately jumps to the start of the except block
and executes the code inside it. When this happens, we say that the
exception has been caught and refer to the code that deals with it as an
exception handler. Statements in the try block after the statement that
raised the exception are not executed.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 243
try:
x = 1/0.3
print 'reciprocal of 0.3 is', x
x = 1/0.0
print 'reciprocal of 0.0 is', x
except:
print 'error: no reciprocal'
(rest of program)
try:
x = 1/0.3
print 'reciprocal of 0.3 is', x
except:
print 'error: no reciprocal'
(rest of program)
1
2
1
3
2
3
4
5
No Error Division By Zero
Figure 12.2: A simple try/except
We can also tell Python what to do when an exception isn’t raised by
adding an else block to the try/except:
Download construct/exceptelse.cmd
>>> def invert(x):
... try:
... i = 1.0 / x
... except:
... print 'caught exception for', x
... else:
... print 'reciprocal of', x, 'is', i
...
>>> invert(1)
reciprocal of 1 is 1.0
>>> invert(0)
caught exception for 0
Here, the statements in the else block are executed only if everything
inside the try block was executed without error. If we were sure we
knew all the places that exceptions could be thrown, we could put these
statements inside the try block. Even then, it is often clearer to put
them in an else block so that people reading the code can see that these
statements are to be executed only if the try block completed normally.
Exception Objects
When Python raises an exception, it creates an object to hold informa-
tion about what went wrong. Typically, this object will contain an error
message, along with a filename and line number to help the program-
mer pinpoint the problem.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 244
Different kinds of errors raise exceptions of different types, which al-
lows a program to choose which kinds of errors to handle by specifying
an exception type in the except. For example:
Download construct/excepttype.cmd
>>> values = [-1, 0, 1]
>>> for i in range(4): # one more than len(values)
... try:
... r = 1.0 / values[i]
... print 'reciprocal of', values[i], 'at', i, 'is', r
... except IndexError:
... print 'index', i, 'out of range'
... except ArithmeticError:
... print 'unable to calculate reciprocal of', values[i]
reciprocal of -1 at 0 is -1.0
unable to calculate reciprocal of 0
reciprocal of 1 at 2 is 1.0
index 3 out of range
Here, the first except block handles only indexing errors; arithmetic
errors (such as dividing by zero) are handled by the second block. If
we want to know exactly what went wrong, we must modify the except
statement so that Python knows which variable to assign the exception
object to:
Download construct/exceptobj.cmd
>>> values = [-1, 0, 1]
>>> for i in range(4): # one more than len(values)
... try:
... r = 1.0 / values[i]
... print 'reciprocal of', values[i], 'at', i, 'is', r
... except IndexError, e:
... print 'error:', e
... except ArithmeticError, e:
... print 'error:', e
reciprocal of -1 at 0 is -1.0
error: float division
reciprocal of 1 at 2 is 1.0
error: list index out of range
Python tests except blocks in order. Whichever matches first gets to
handle the exception. The order matters because exceptions are ar-
ranged in a hierarchy, as shown in Figure 12.3, on the next page. This
allows a program to handle ZeroDivisionErrors one way and all other Arith-
meticErrors another, but only if the handler for the first comes before
the handler for the second. An except that doesn’t specify an exception
class catches everything, so if there is one, it must appear last, like the
else in an if-elif-else chain.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 245
ZeroDivisionError
ArithmeticError
IndexError
Exception
StandardError
LookupError
Figure 12.3: Exception hierarchy
Functions and Exceptions
What if Python can’t find a matching exception handler in a particu-
lar try/except block? For example, what happens when Python tries to
divide by zero in the following code?
Download construct/uncaught.py
values = [-1, 0, 1]
for i in range(4): # one more than len(values)
try:
r = 1.0 / values[i]
print 'reciprocal of', values[i], 'at', i, 'is', r
except IndexError, e:
print 'error:', e
The answer is that Python keeps a stack of exception handlers, just
like its stack of function calls (see Figure 12.4, on the following page).
When an exception is raised, Python takes handlers off the exception
handler stack one by one until it finds a handler that matches and then
executes it. This means that the code to handle an exception can be a
long way away from the place where the exception occurred. It also
means that one exception handler can take care of exceptions from
many pieces of code.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 246
ArithmeticError?
def average(nums): sum = 0.0 for n in nums: sum += n return sum / len(nums)
try: average([])except IndexError: print 'Index out of range'except ArithmeticError: print 'Error with calculations'
IndexError?
ZeroDivisionError
__main__
average
Figure 12.4: The exception handler stack
For example, suppose we are using several functions that calculate
statistics on numbers stored in lists. Some of those functions might
raise exceptions because of division by zero. We don’t particularly care
which function call is at fault, so we can deal with them all at once like
this:
Download construct/exceptfunc.cmd
def average(nums):
sum = 0.0
for n in nums:
sum += n
return sum / len(nums)
...other function definitions skipped...
values = read_values_from_file()
try:
print 'average:', average(values)
print 'median:', median(values)
print 'standard deviation:', std_dev(values)
except ArithmeticError:
print 'Error in calculations'
As you can see, the code that handles the error is outside the functions
that are doing the work. This means that the functions can focus on
doing their jobs and let whoever is calling them worry about how to
handle mistakes.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 247
Raising Exceptions
You don’t have to wait for Python to notice that something has gone
wrong—you can raise an exception yourself using the raise keyword,
followed by the type of exception that you want to raise. When you do
this, it’s good style to include an error message to explain what went
wrong:
Download construct/raise.cmd
>>> def divide(top, bottom):
... if bottom == 0:
... raise ValueError('divisor is zero')
... else:
... return top / bottom
...
>>> for i in range(-1, 2):
... try:
... print divide(1, i)
... except ValueError, e:
... print 'caught exception for', i
...
-1
caught exception for 0
1
You will still sometimes see the raise statement written like this:
raise ValueError, 'divisor is zero'
However, that style is now deprecated; that is, Python’s creator has said
that it will be removed from the language one day, so new programs
shouldn’t use it (and old ones should be fixed).
Exceptional Style
When working with exceptions, you should always follow two rules.
First, always indicate errors in functions by raising exceptions, rather
than returning None, -1, False, or some other “special” value.1 This adds
clarity by separating error-handling code from “normal” code; in addi-
tion, experience shows that, sooner or later, you are going to want some
of your functions to return those special values, at which point you will
have no way of distinguishing success from failure. Alternatively, you
1. Note that Python’s own str.find breaks this rule by returning -1 if something can’t be
found instead of throwing an exception. In this case, practicality beats purity. Returning
-1 is only slightly more efficient than raising an exception, but substring searches are so
common that Python’s designers decided it was worth it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXCEPTIONS 248
will need to completely rewrite the code so that it both raises exceptions
and returns the appropriate values.
The second rule is often phrased as, “Throw low, catch high.” This
means that you should throw lots of very specific exceptions in the
lower levels of your program but catch them only at a higher level in the
exception hierarchy where you can take corrective action. For example,
instead of writing code like this:
def first():
try:
...body of first function...
except Exception, e:
print 'error occurred', e
def second():
try:
...body of second function...
except Exception, e:
print 'error occurred', e
if __name__ == '__main__':
first()
second()
a programmer should write this:
def first():
...body of first function...
def second():
....body of second function...
if __name__ == '__main__':
try:
first()
second()
except Exception, e:
print 'error occurred', e
Handling errors in a small number of places like this makes it easy to
change the way they are handled. If, for example, you decide to write
application errors to a log file as well as print them to the screen, you
have to change only one piece of code in the second example. Similarly,
if this program were made part of a larger application with a graphical
interface that required errors to be displayed in dialogs rather than
on the command line, there would be exactly one place to make the
change.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 249
12.3 Testing
Now that we have looked at how to deal with errors, we can discuss
how to test programs so that they don’t happen. The first thing to point
out is that testing on its own is not enough to produce high-quality
software. As Steve McConnell said in [McC04], “Trying to improve the
quality of software by doing more testing is like trying to lose weight by
weighing yourself more often.” Every time you sit down to test a piece
of software, you should therefore start by asking yourself a few simple
questions: What kind of problem am I looking for? What’s the most
efficient way to find out if the problem exists?
Terminology
Let’s start by defining some terms. A functional test looks at the behav-
ior of the system as a whole, just as its eventual users will. A unit test
exercises just one isolated component of a program. For example, here’s
a simple-minded unit test of a function that compares two strings and
returns -1, 0, or 1 if the first comes before, is the same as, or comes
after the second alphabetically:
Download construct/compare_strings.py
if my_compare('abc', 'def') == -1:
print 'success'
else:
print 'failure'
In Nose, this would be as follows:
Download construct/compare_strings_nose.py
import nose
def test_compare():
'''Test comparison of unequal strings.'''
assert my_compare('abc', 'def') == -1
if __name__ == '__main__':
nose.runmodule()
Two more frequently used terms are black-box testing and glass-box
testing. As the name suggests, black-box testing considers only the
inputs and outputs of the code being tested, not its implementation.
(The term is borrowed from electrical engineering, where a black box is
often the thing being tested.) In glass-box testing, on the other hand,
the tester is able to look inside the program and see how it works.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 250
Glass-box testing may sound more effective, since the tester is able to
see which parts of the code to stress. In practice, though, knowing how
something is implemented often leads to psychological bias. Thus, it is
therefore often better to treat the code the same way its eventual users
will and test it by comparing its actual behavior to its specification.
Regression testing is another term that comes up fairly often. It is the
practice of rerunning tests to make sure that recent changes haven’t
broken anything. Regression testing is practical only if the tests are
automated and if changes to program behavior are relatively infrequent.
If the first condition doesn’t hold, then human beings must retest man-
ually every time the program changes. This is boring, expensive, and
error-prone, so in practice, most projects don’t do it.
The second condition is a corollary of the first. If the program’s behavior
is changing on a daily basis, then the project’s automated tests have
to be rewritten every day as well, which cancels out the advantages of
automation. This is one of the main reasons why it’s important to spend
time up front on design before writing code.
Unit Testing
For now we will focus on unit testing, since that’s what programmers
do while they’re programming. If you would like to know more, [AW06,
Whi03, WT04] are practical guides on what to look for in desktop and
web applications and how to find security holes.
As discussed in Section 4.5, Testing, on page 70, a unit test can have
one of three outcomes: pass, fail, or error. To classify a result, though,
you must know how the code being tested is supposed to behave; that
is, you must have some kind of specification. The more precise that
specification is, the easier it will be to determine whether the code is
doing what it’s supposed to do.
Independence
For unit testing to be useful, every test must be independent; in other
words, the outcome of one test must not depend on the outcome of
other tests. There are two reasons for this. First, if this rule is violated,
then every time we change a test that runs early, we have to check all
subsequent tests to see whether they have been affected. The second,
and more important, reason is that failures in early tests can hide bugs
that later tests are supposed to find.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 251
For example, try running this code:
Download construct/test_side_effects.py
import nose
numbers = [1, 3, -1, 5]
def test_max():
assert max(numbers) == 5
def test_min():
assert min(numbers) == -1
def test_append():
old_len = len(numbers)
numbers.append(99)
assert len(numbers) == old_len + 1
if __name__ == '__main__':
nose.runmodule()
If Nose decides to run the tests in the order they’re given, everything
will be OK. But if Nose runs the tests in a different order or if someone
decides to put the tests in alphabetical order, test_max will suddenly
start to fail, because test_append is changing the values it is being run
on.2 As a rule, therefore, the data structures manipulated in the test,
called its fixture, should be used only once.
Limitations
Now for the bad news. In practice, it’s practically impossible to com-
pletely test any nontrivial piece of software. Suppose that you have a
function that compares two seven-digit phone numbers and returns
True if the first is greater than the second. There are (1072
) possible
inputs; if you can compare 100 million pairs of numbers per second, it
will take 15.5 days to check all combinations. If the inputs are seven-
character alphabetic strings, rather than numbers, it will take a little
more than twenty-five years, and that’s just the first function—your
program probably contains dozens, or thousands, many of which will
be much more complex.
2. Of course, part of the problem is the use of the global variable numbers. If each test
function used a local variable, it would be safe for test_append to its list.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 252
Luckily, testing isn’t as futile as this analysis makes it sound. In prac-
tice, it’s usually enough to test the following:
• Boundary cases, such as sorting an empty list or list containing a
single value
• The Simplest interesting cases, such as sorting lists containing two
values (one where the items are in order and one where they are
not)
• The General cases, such as sorting a longer list
If your code works for these cases, there’s a good chance it will work for
all other cases as well.
Of course, you should also test cases that are expected to fail, since
error handling is part of a function’s specification too. Test with invalid
input or too little input or too much, and make sure it fails the way the
specification says it should. Finally, do a few sanity checks. If informa-
tion appears in two or more places, for example, check that all occur-
rences are still consistent after the action. For example, if a program
stores both a list of numbers and their sum, make sure that the sum
is still correct after any operation that manipulates the list.
Above all, remember that human beings are creatures of habit; we tend
to make the same kinds of errors over and over again. So, good testers
will test for common errors first, but great ones build catalogs of errors
to refer back to whenever they test a new piece of code. Here are some
tests that often turn up problems in code:
• For numbers: Zero; the largest and smallest allowed numbers; one
less and one more than the largest and smallest values.
• For data structures: Empty structures; structures containing ex-
actly one element; structures containing the maximum allowed
number of elements; structures containing duplicates (for exam-
ple, a letter appears three times in a string); structures containing
aliased values (for example, a sublist appears three times in an
outer list).
• For searching: No match found; one match found; multiple
matches found; everything matches.
Over time, you’ll commit fewer and fewer errors as you catalog them
and, subsequently, become more conscious of them. And that’s really
the whole point of focusing on quality. The more you do it, the less
likely it is for problems to arise.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TESTING 253
Test-Driven Development
Since the late 1990s, a style of programming called test-driven devel-
opment (TDD) has become increasingly popular. The idea behind it
is that tests are actually specifications. Each test says, “Given these
inputs, this code should produce this result.” Therefore, programmers
who practice TDD write their test cases first and then write the code to
be tested.
This sounds backward, but studies have shown that it can actually
make programmers more productive, for several reasons:
• Tests provide a more precise specification than sentences.
• TDD gives programmers a finish line to aim for: a function is fin-
ished when its unit tests all pass.
• TDD ensures that tests actually get written, since even the best
programmers are often too tired or too rushed to write a good set
of tests after the code appears to be working.
• Writing tests first gives programmers a chance to “test-drive” the
code they’re about to write before actually writing it. For example,
if a programmers finds that it’s awkward to create tests for the
matrix functions she’s about to write, she can change the interface
before it is set in stone.
As an example of TDD, suppose we need to write a function that cal-
culates a running sum of the values in a list. For example, if the list is
[1, 2, 3], the result should be [1, 1+2, 1+2+3], which is [1, 3, 6]. This seems
simple enough, but look at what’s missing: it’s not clear if a new list
should be created or the old one should be overwritten, and there’s no
mention of how to handle errors.
Now look at this:
Download construct/running_sum_tdd.py
import nose
def test_empty_list():
assert running_sum([]) == []
def test_single_value():
assert running_sum([1]) == [1]
def test_two_values():
assert running_sum([1, 3]) == [1, 4]
def test_three_values():
assert running_sum([1, 3, 7]) == [1, 4, 11]
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEBUGGING 254
def test_negative_values():
assert running_run([-1, 1]) == [-1, 0]
def test_mixed_types():
assert running_sum([1, 3.0]) == [1, 4.0]
def test_string():
try:
running_sum('string')
assert False
except ValueError:
pass
except:
assert False
def test_non_numeric():
try:
running_sum(['string'])
assert False
except ValueError:
pass
except:
assert False
Even without docstrings, these tests tell us that the function
• has to be able to handle a mix of integer and floating-point values,
• doesn’t have to add things that aren’t numbers, and
• should produce a ValueError exception when given invalid input.
That’s a pretty good specification.
12.4 Debugging
As most programmers have discovered, debugging3 is what you get to
do when you didn’t do enough testing. Tracking down and eliminating
bugs in your programs is part of every programmer’s life, so this chapter
introduces some techniques that can make debugging more efficient
and give you more time to do the things you’d rather be doing.
Debugging a program is like diagnosing a medical condition. To find
the cause, you start by working backward from the symptoms (or, in a
3. In 1945, Admiral Grace Hopper’s team “debugged” Harvard University’s Mark II Aiken
Relay Calculator by removing a moth trapped in one of the relays. However, the term dates
back to at least Thomas Edison’s time.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
DEBUGGING 255
program, its incorrect behavior), then come up with a solution, and test
it to make sure it actually fixes the problem.
At least, that’s the right way to do it. Many beginners make the mis-
take of skipping the diagnosis stage and trying to cure the program by
changing things at random. Renaming a variable or swapping the order
in which two functions are defined might actually fix the program, but
millions of such changes are possible. Trying them one after another in
no particular order can be an inefficient waste of many, many hours.
Here are some rules for tracking down the cause of a problem:
1. Make sure you know what the program is supposed to do. Some-
times this means doing the calculation by hand to see what the
correct answer is. Other times, it means reading the documenta-
tion (or the assignment) carefully or writing a test.
2. Repeat the failure. You can debug things only when they go wrong,
so find a test case that makes the program fail reliably. Once
you have one, try to find a simpler one; doing this often provides
enough clues to allow you to fix the underlying problem.
3. Divide and conquer. Once you have a test that makes the program
fail, try to find the first moment where something goes wrong.
Examine the inputs to the function or block of code where the
problem first becomes visible. If they are wrong, look at how they
were created, and so on.
4. Change one thing at a time, for a reason. Replacing random bits of
code on the off-chance they might be responsible for your prob-
lem is unlikely to do much good. (After all, you got it wrong the
first time....) Each time you make a change, rerun your test cases
immediately.
5. Keep records. After working on a problem for an hour, you won’t
be able to remember the results of the tests you’ve run. Like any
other scientist, you should keep records. Some programmers use a
lab notebook; others keep a file open in an editor. Whatever works
for you, make sure that when the time comes to seek help, you
can tell your colleagues exactly what you’ve learned.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PATTERNS 256
12.5 Patterns
It’s hard to figure out what a program is supposed to do if it is cluttered
or disorganized. Good design is therefore the most important program-
ming tool of all.
One of the keys to organizing programs is to recognize that variables are
usually used in stereotypical ways. Learning to recognize these variable
roles helps us read code more quickly; it also helps us figure out when
something is going wrong. The following sections, taken from [KS04],
describe some typical roles.
Fixed Values
As the name suggests, a fixed value is a variable whose value doesn’t
change once it is assigned. Fixed values are often used to give human-
readable names to “magic numbers” like 9.81 (Earth-normal gravity in
meters per second squared) or 3.14159 (a rough approximation to pi).
By convention, they are placed at the top of the program, as in the
following example:
Download construct/fixedvalue.cmd
>>> SECONDS_PER_DAY = 24 * 60 * 60
>>> instant = 10**3
>>> while instant <= 10**7:
... print "%10d seconds is %8.2f days" % \
... (instant, (1.0 * instant / SECONDS_PER_DAY))
... instant *= 10
1000 seconds is 0.01 days
10000 seconds is 0.12 days
100000 seconds is 1.16 days
1000000 seconds is 11.57 days
10000000 seconds is 115.74 days
As this example shows, they are also often given uppercase names to
make them stand out from “normal” values.
Stepper and Counter
A stepper is a variable that “steps through” a sequence of values in
some predictable way. The index variables in for loops are almost always
steppers; so is the time variable in the previous example. A stepper
is very much like a counter, but the latter term normally refers to an
integer that keeps track of how many things have been seen so far.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PATTERNS 257
Most-Wanted Holder
Normally, we search for a value in a collection by checking each value in
turn and keeping a reference to the best one seen so far. That reference
is called a most-wanted holder, since it holds the value we want most.
For example, here’s a function that searches for the largest value in a
list that is less than some threshold:
Download construct/largethresh.py
def largest_below_threshold(values, threshold):
'''Find the largest value below a specified threshold. If no value is
found, returns None.'''
result = None
for v in values:
if v < threshold:
result = v
break
if result is None:
return None
for v in values:
if result < v < threshold:
result = v
return result
Most-Recent Holder
A most-recent holder holds the value most recently seen from some
sequence of values. In a sense, a stepper is just a special case of a
most-recent holder. The only distinction is that the stepper is step-
ping through regular values (like the elements of a list), while a most-
recent holder is assigned values that can come from anywhere. Here’s
an example:
Download construct/mostrecent.py
# A really simple number-guessing game
while True:
input = raw_input('Enter a number: ')
try:
val = int(input)
except ValueError:
print 'I said, a *number*...'
if val == 7:
print 'You guessed it!'
break
else:
print 'Try again'
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PATTERNS 258
Since users can type in anything they want, this program validates the
input (in other words, checks that it’s at least plausible) before doing
anything else with it.
Container
A container is a value that holds other values. Lists are clearly con-
tainers, but what about strings? They can be containers too, if we find
ourselves adding or removing characters:
Download construct/container.cmd
>>> whole = 'selenium'
>>> for i in range(len(whole)/2):
... print whole
... whole = whole[1:-1]
selenium
eleniu
leni
ni
This example shows that patterns like “stepper” and “container” are a
matter of interpretation, rather than part of the program itself. In the
previous example, the string "selenium" isn’t being modified in place—
that’s not allowed in Python. However, when we’re writing the program,
it’s helpful to think of it that way. So, is whole a container of characters
or a most-recent holder?
Gatherer
A gatherer (sometimes called an accumulator) collects individual values.
For example, the variable result in the following code is a gatherer; it
collects values by adding them:
Download construct/gatherer.py
def add_up(values):
'''Return the sum of the values in a list, or 0 if the list is empty.'''
result = 0
for sum in values:
result += sum
return result
Temporary
A temporary is a variable that exists only for a short time. Temporaries
are typically used to store partial results in order to make code easier
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
PATTERNS 259
to read. For example, here’s a function that returns the two roots of a
quadratic equation:
Download construct/roots.py
from math import sqrt
def roots(a, b, c):
'''Return real roots of a quadratic, or None.'''
temp = b**2 - 4*a*c
if temp < 0:
return None
temp = sqrt(temp)
left = (-b + temp) / (2 * a)
right = (-b - temp) / (2 * a)
return left, right
Compare this with a version that uses fewer lines but doesn’t use a
temporary:
Download construct/roots_full.py
from math import sqrt
def roots(a, b, c):
'''Return real roots of a quadratic, or None.'''
if sqrt(b**2 - 4*a*c) < 0:
return None
left = (-b + sqrt(b*2 - 4*a*c)) / (2 * a)
right = (-b - sqrt(b**2 - 4*a*c)) / (2 * a)
return left, right
Did you spot the bug in the second version? Code that seems to be
repeated is never read as carefully as code that isn’t; using temporaries
to store intermediate results therefore reduces the chances of a lurking
bug.
One-Way Flag
A one-way flag is a variable whose value changes just once to show
that something has occurred.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 260
For example, this function finds the lines in a data file that were added
after a particular date:
Download construct/afterdate.py
def after_date(input_file, date):
'''Return the lines that were added to a file after a certain date.'''
keep_it = False
result = []
for line in input_file:
if keep_it:
result.append(line)
elif get_date(line) >= date:
keep_it = True
return result
The keep_it variable is initially set to False to signal that we haven’t yet
seen a line in the file added after the desired date. As soon as we do see
that line, the flag becomes True.
12.6 Summary
In this chapter, we learned the following:
• Specifying default values for parameters can make functions eas-
ier to use; letting them take variable-length parameter lists can
make them more flexible.
• When an error occurs in a program, the program should raise an
exception with a meaningful type and an informative error mes-
sage. Programs should catch exceptions only in a small number of
places where they can do something to report or correct the error.
• Functional testing looks at the program as a whole, while unit
testing focuses on its components, and regression testing checks
whether things that used to work still do.
• Unit tests should operate independently so that an error in one
will not affect the results reported for the rest.
• The most cost-effective way to test is to focus on boundary cases.
It’s also important to test that software fails when it’s supposed to.
• Test-driven development consists of writing unit tests first and
then writing code to make those tests pass. TDD helps program-
mers be more efficient by getting them to focus on what code is
supposed to before worrying about how it is going to work.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 261
• To debug software, you have to know what it is supposed to do
and be able to repeat the failure. Simplifying the conditions that
make the program fail is an effective way to narrow down the set
of possible causes.
• Most variables in programs are used in one of a small number
of stereotyped ways. Knowing these patterns helps with design,
debugging, and communication.
12.7 Exercises
Here are some exercises for you to try on your own:
1. Python allows programmers to pass values to functions in any
order they like, provided they specify the parameters’ names. For
example:
>>> def byName(red, green, blue):
... print red, green blue
...
>>> byName(green=0.1, blue=0.5, red=0.4)
0.4 0.1 0.5
Why is this useful? Why is it dangerous? Explain why you do
or don’t think that Python (or any other programming language)
should allow this or not.
2. The total function in Section 12.1, Default Parameter Values, on
page 237 uses None as a default value for end. Why is this more
sensible than using an integer like -1?
3. Python actually provides two ways to create variable parameter
lists. The first, discussed in Section 12.1, Variable Parameter Lists,
on page 239, takes all the “extra” values passed to the function
call, puts them into a tuple, and assigns that tuple to a parameter
with a star (*) in front of its name. The second takes any extra
named parameters, puts them in a dictionary, and assigns it to a
parameter with two stars (**) in front of its name.
a) Find the description of this feature in the official Python doc-
umentation at http://www.python.org and give its URL.
b) Write a short example to show how this feature works, along
with 100–150 words of explanation.
c) When should you use this feature, and why?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 262
4. Your job is to come up with tests for a function called line_intersect,
which takes two lines as input and returns their intersection.
More specifically:
• Lines are represented as pairs of distinct points, such as [[0.0,
0.0], [1.0, 3.0]].
• If the lines do not intersect, line_intersect returns None.
• If the lines intersect in one point, line_intersect returns the
point of intersection, such as [0.5, 0.75].
• If the lines are coincident (that is, lie on top of one another),
the function returns its first argument (that is, a line).
What are the six most informative test cases you can think of?
(That is, if you were allowed to run only six tests, which would tell
you the most about whether the function was implemented cor-
rectly?) Write out the inputs and expected output of these six tests,
and explain why you would choose them. Do not worry about error
handling (yet).
5. Your next job is to test whether line_intersect handles errors cor-
rectly. You have been told that it is supposed to raise a ValueError
exception if either argument is not a pair or pairs of floating-point
numbers and that it is supposed to raise a GeometryError excep-
tion if the two points used to define a line are not distinct (that is,
if something like [1.0, 1.0], [1.0, 1.0]] is given as an input). What six
tests would you write first to check that line_intersect was handling
errors correctly? In a sentence or two, explain whether you think
that testing error handling is more important than testing that the
function gives the correct answer for valid input, or vice versa.
6. A friend of yours has written a function called common_name that
takes two strings as input. The first is the one- or two-letter sym-
bol for a chemical element, such as "He" or "Fe". The second is the
two-letter international code for a language, such as "EN" (English)
or "DE" (Deutsch, that is, German). You have been told that the
function handles 106 elements and 26 languages. If you don’t
want to test every element with every language, then how many
different inputs do you have to give it to feel confident that it works
correctly? What assumptions are you making about those inputs
in doing your calculation?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 263
7. Another friend of yours has written a function called average,
which returns the average of a list of floating-point numbers. She
has tested it as follows:
Download construct/testfloat.py
from math import abs
from mylibrary import average
test_cases = [
[0.0, [ 0.0]],
[0.0, [-1.0, 1.0]],
[1.0, [ 0.0, 2.0]],
[2.0, [ 0.0, 1.0, 2.0, 3.0, 4.0]],
....
]
passes = failures = 0
for (expected, values) in test_cases:
actual = average(values)
if actual == expected:
passes += 1
else:
failures += 1
print 'passes:', passes
print 'failures:', failures
She has implemented average correctly, but several of her test
cases are failing. Explain why; identify which single line of code
she should change, and how to change it, to improve her tests.
8. A function called is_salt takes the chemical symbols of two elements
as input and returns True if the combination of those two elements
is a salt (in other words, a combination of an acid and a base).
You happened to know that the symbols for the elements that are
considered acids and bases are contained in sets called Acids and
Bases, respectively. A friend of yours thinks that an easy way to
test is_salt would be to simplify the Acids and Bases sets and then
run the tests:
Download construct/testsalt.py
Bases = set(["Mg"]) # nothing but magnesium
Acids = set(["Cl"]) # nothing but chlorine
def test_MgCl():
'''Works if the base is a base, and the acid is an acid.'''
assert is_salt("Mg", "Cl")
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 264
def test_LiCl():
'''Doesn't work because the 'base' isn't a base.'''
assert not is_salt("Ar", "Cl")
def test_MgO():
'''Doesn't work because the 'acid' isn't an acid.'''
assert not is_salt("Mg", "O")
def test_neither():
'''Doesn't work because neither is right.'''
assert not is_salt("Ar", "O")
There are (at least) two things wrong with this idea. Briefly explain
what they are.
9. Using Nose, write four tests for a function called all_prefixes in a
module called testprefixes.py that takes a string as its input and
returns the set of all nonempty substrings that start with the first
character. For example, given the string "lead" as input, all_prefixes
would return the set {"l", "le", "lea", "lead"}. Include a comment ex-
plaining what crucial piece of information is missing from the
specification for all_prefixes that prevents you from testing it fully.
10. Using Nose, write the ten most informative tests you can think of
for a function called is_sorted in a module called testsorting.py that
takes a list of integers as input and returns True if they are sorted
in nondecreasing order, and False otherwise. If the argument is not
a list or if any of the values are not integers, the function should
raise a ValueError exception. Include a short comment justifying the
number of tests you chose to write.
11. The function remove_all takes two arguments: a list and an item
you want to completely remove from the list. If the item occurs in
the list multiple times, every instance of it is removed. If the item
does not occur in the list, the list remains unmodified.
Explain what is wrong with the following tests for the function
remove_all:
Download construct/remove_all.py
empty_list = []
one_item_list = ['He']
multi_item_list = ['Ne', 'Ar', 'He', 'He']
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 265
def test_remove_from_empty():
remove_all(empty_list, 'He')
assert len(empty_list) == 0
def test_remove_from_one_item_list():
remove_all(one_item_list, 'He')
assert len(one_item_list) == 0
def test_remove_something_else():
remove_all(one_item_list, 'Pb')
assert len(one_item_list) == 1
def test_remove_multiple():
remove_all(multi_item_list, 'He')
assert len(multi_item_list) == 2
12. At first glance, it seems that the only way to test a function that
reads from a file is to store some sample files on disk and have the
function open and read those files. However, the standard Python
library includes a module called StringIO (and a faster version called
cStringIO) that makes file I/O easier to test. Explain how in one
short paragraph, and then use cStringIO to test a function called
count_lines (taken from a module called file_util).
13. The following function is broken. The docstring describes what it’s
supposed to do:
Download construct/minmaxexercise.py
def find_min_max(values):
'''Print the minimum and maximum value from the given collection of
values.'''
min = None
max = None
for value in values:
if value > max:
max = value
if value < min:
min = value
print 'The minimum value is %s' % min
print 'The maximum value is %s' % max
What does it actually do? Use a debugger to verify. What line do
you need to change to fix it?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 266
14. The following function is also broken but only under certain cir-
cumstances:
Download construct/summationexercise.py
def summation(limit):
'''Return the sum of the numbers from 0 to limit.'''
total = 0
current = limit
while current != 0:
total += current
current -= 1
return total
a) What type of input causes the function to fail?
b) How does the function behave during a failure? Use a debug-
ger to step through what happens.
c) What do you need to change to fix it?
15. Under what circumstances will the following function throw an
exception? Use try and except to ensure the function returns the
correct value even when an exception occurs.
Download construct/reciprocalexercise.py
def compute_reciprocals(values):
'''Return a list of the reciprocals of the given list of values.
If a value has no reciprocal, it will be assigned a value of
None in the returned list.'''
reciprocals = []
for value in values:
reciprocals.append(1 / value)
return reciprocals
16. For each function, identify the role of the variables v0, v1, and so
on. Also, give each poorly named variable a better name.
Download construct/roleexercise1.py
def find_last(filename, string):
v0 = 0
v1 = [None, None]
v2 = open(filename, "r")
for v3 in v2:
v0 += 1
if string in v3:
v1 = [v0, v3]
return v1
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 267
Download construct/roleexercise2.py
from math import sqrt
def standard_deviation(values):
v0 = 0.0
for v1 in values:
v0 += v1
v2 = v0 / len(values)
v3 = 0.0
for v4 in values:
v3 += (v4 - v2) ** 2
v5 = v3 / len(values)
return sqrt(v5)
17. Write three functions that compute the mean, median, and mode,
respectively, of a list of values. For each local variable you use,
indicate its role using comments. Don’t forget to identify the role
of variables you use in loops. Use the following to get started:
Download construct/roleexercise3.py
def mean(values):
'''Return the arithmetic mean of the list of values; i.e. the
sum of all the values divided by the total number of values.
If the list contains non-numeric elements, this function throws
a ValueError.'''
def median(values):
'''Return the median of the list of values. For an odd number
of values, if the numbers are sorted, the median occurs at the
center of the list. For an even number of values, the median is
the mean of the two middle values in the sorted list of values.'''
def mode(values):
'''Return a list containing the mode of the list of values. The
mode is the most frequently occurring values in the list. For example,
the mode of [0, 1, 0, 3, 2, 4, 1] is [0, 1].'''
18. Occasionally programmers don’t close files that they have opened.
This is bad style. The computer’s operating system allows a run-
ning program to have only a limited number of files open at any
time, so if programs don’t close files they’re done with, they will
eventually not be able to open new ones. Modify the following
function so that it always closes the file it has opened, even if
something goes wrong while the file’s contents are being read.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 268
Download construct/add_error_handling.py
def process_file(filename):
'''Read and print the contents of a file.'''
f = open(filename, 'r')
for line in f:
line = line.strip()
print line
19. Take another look at the function we wrote in Section 8.4, Multiline
Records, on page 177 to read molecule descriptions from a file:
Download fileproc/multimol_2.py
def read_molecule(r):
'''Read a single molecule from reader r and return it,
or return None to signal end of file.'''
# If there isn't another line, we're at the end of the file.
line = r.readline()
if not line:
return None
# Name of the molecule: "COMPND name"
key, name = line.split()
# Other lines are either "END" or "ATOM num type x y z"
molecule = [name]
reading = True
while reading:
line = r.readline()
if line.startswith('END'):
reading = False
else:
key, num, type, x, y, z = line.split()
molecule.append((type, x, y, z))
return molecule
This function assumes that every file is correctly formatted. In
the real world, data files often aren’t. Lines may be missing or
in the wrong order, values may be out of order on a line, and so
on. In keeping with the philosophy of defensive programming, add
assertions to this function so that it fails with appropriate error
messages for badly formatted input.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 269
20. Consider the following function:
def findButNotAfter(source, prefix, pattern):
'''Return the index of the first occurrence of 'pattern' in
'source' that is NOT immediately after an occurrence of
'prefix'. For example, findButNotAfter('abcdcd', 'ab', 'cd')
returns 4, since the first occurrence of 'cd' comes immediately
after an 'ab', but findButNotAfter('abxcdcd', 'ab', 'cd')
returns 3.'''
a) What ambiguities or omissions are there in this specification?
b) Rewrite the function’s docstring to address the shortcomings
you identified earlier.
c) Write at least ten Nose tests to check that the function meets
your revised specification.
d) Write a body for findButNotAfter that passes all of your tests.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 13
Object-Oriented ProgrammingAs you learned in Section 2.3, What Is a Type?, on page 22, a type
is a set of values along with the operations that can be performed on
those values. For example, the type bool consists of the values True and
False, plus the operations and, or, and not, while the type int has approx-
imately 232 values and a wide range of operations, such as addition and
subtraction.
Types like bool and int are built into programming languages because
they are useful in almost every program. Thousands of other types,
from File to Nematode, are just as useful, but not to as many people;
trying to include all of them would be a never-ending task and would
make the language so large that it would be practically impossible to
work with.
One of the greatest advances in computer science was the realization
that allowing programmers to define new types of their own can make
programs easier to write and understand. In almost all modern lan-
guages, including Python, programmers do this by defining a class and
then creating objects that are instances of the class. You can think of
a class as being like a species, like Pygoscelis antarctica (the chinstrap
penguin); its instances are then particular penguins.
Languages working this way are (unsurprisingly) called object-oriented.
This chapter will show you how to create object-oriented programs in
Python and why you would want to write programs that way. We will
start by showing how to create a simple class, then look at some of the
theory behind object-oriented programming, and finally examine some
of the “extras” Python provides to make programmers’ lives easier.
Prepared exclusively for Trieu Nguyen
CLASS COLOR 271
13.1 Class Color
To explain classes, over the next few sections we will build a Color class
of our own that does the same things as the RGB model described in
Section 4.4, Pixels and Colors, on page 68. Let’s start with a very simple
(but not very useful) class definition:
Download oop/color_simple.py
class Color(object):
'''An RGB color, with red, green and blue components.'''
pass
Just as the keyword def tells Python that we’re defining a new function,
the keyword class signals that we’re defining a new type. The (object)
part says that class Color is a kind of object;1 the docstring describes
the features of a Color object, while pass says that the object is blank—
that is, it doesn’t store any data and doesn’t offer any new operations.
The following sections will add data and operations to Color to make
it more useful. Even without that, though, we can make a new Color
object and assign it to a variable called black:
Download oop/color_obj.cmd
>>> black = Color()
Using the name of the class as if it were a function tells Python to
find the class and build one instance of it. We can check that this has
worked by printing the value of black:
Download oop/color_show_raw.cmd
>>> black
<__main__.Color object at 0xb7dbd24c>
That tells us that an object has been created (and where it happens to
be in memory), but nothing more. To make the object more useful, we
need to attach some red, green, and blue values to it. One way to do
this—a very bad way, which we will clean up in a couple of sections—is
simply to assign the values we want:
Download oop/color_add_values.cmd
>>> black.red = 0
>>> black.green = 0
>>> black.blue = 0
1. This is needed for historical reasons, since early versions of Python did things a
slightly different way.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CLASS COLOR 272
The variable black now refers to a Color object that has three instance
variables called red, green, and blue. To prove that they’re there, we can
examine their values:
Download oop/color_print_parts.cmd
>>> black.red
0
>>> black.green
0
>>> black.blue
0
An object’s instance variables are like the local variables in a function
call. Just as local variables can be accessed only inside their func-
tion and last only as long as the function call, instance variables can
be accessed (using dot notation) only through the object that contains
them and live only as long as that object.
Another way to think about objects is to compare them to dictionaries.
We could store these same RGB values using this:
black = {"red" : 0, "green" : 0, "blue" : 0"}
and then access them using black["red"] instead of black.red. As we’ll see,
though, objects allow us to do things that dictionaries can’t.
Methods
Before showing how instance variables ought to be created, let’s take
a look at how we can add operations to a class. Suppose, for example,
that we want to calculate the lightness of a color, that is, how close it is
to being pure white. By definition, a color’s lightness is the average of
its strongest and weakest RGB values scaled to lie between 0 and 1. As
a function, this is as follows:
Download oop/color_lightness_func.py
def lightness(color):
'''Return the lightness of color.'''
strongest = max(color.red, color.green, color.blue)
weakest = min(color.red, color.green, color.blue)
return 0.5 * (strongest + weakest) / 255
If we want to make the calculation a method of the class Color, we
simply move its definition into the class.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CLASS COLOR 273
purple
__main__
self
Color.lightness
>>> purple = Color()
>>> purple.lightness()red
green
blue
Color
Runtime stack after
calling purple.lightness()
Figure 13.1: Automatic self
Download oop/color_lightness_method.py
class Color(object):
'''An RGB color, with red, green and blue components.'''
def lightness(self):
'''Calculate the lightness of this color.'''
strongest = max(self.red, self.green, self.blue)
weakest = min(self.red, self.green, self.blue)
return 0.5 * (strongest + weakest) / 255
Actually, we do one other thing as well. We take out the parameter color
and replace it with one called self. Whenever Python called a method
for an object, it automatically passes a reference to that object as the
method’s first argument (see Figure 13.1). This means that when we call
lightness, we don’t need to give it any arguments—Python supplies the
object itself for us. Inside the method, we can then access the object’s
instance variables using the usual dot notation on that variable:
Download oop/color_lightness_call.cmd
>>> purple = Color()
>>> purple.red = 255
>>> purple.green = 0
>>> purple.blue = 255
>>> purple.lightness()
0.5
The rule about self means that when you define a method, you must
include one more parameter than you’re actually going to pass in. Alter-
natively, when you call a method, you provide one less parameter than
the method’s definition seems to require. Forgetting this is a very com-
mon beginner’s mistake. Forgetting to put self in front of something that
you intended to be a member variable is another common beginner’s
mistake.Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CLASS COLOR 274
For example, if the class was defined this way:
Download oop/broken_color_self.py
class Color(object):
'''An RGB color, with red, green and blue components.'''
def lightness(self):
'''Return the lightness of this color.'''
# Fails: no such variables 'red', 'green', and 'blue'
strongest = max(red, green, blue)
weakest = min(red, green, blue)
return 0.5 * (strongest + weakest) / 255
then purple.lightness() would produce an error, because the variables
red, green, and blue don’t exist in lightness, even though purple.red, pur-
ple.green, and purple.blue have been assigned values.
Constructors
Let’s take another look at the color purple:
Download oop/color_purple.cmd
>>> purple = Color()
>>> purple.red = 128
>>> purple.green = 0
>>> purple.blue = 128
We could create more colors this way, but it requires us to type in the
name of the same variable over and over. What is worse—far worse—is
that if the user forgets to add all the right instance variables, then the
object’s methods will probably fail. For example, suppose we created
the color yellow:
Download oop/yellow_create_error.cmd
>>> yellow = Color()
>>> yellow.red = 128
>>> yellow.green = 128
That looks all right. Yellow is equal parts red and green, and both values
have been set. However, if we try to call lightness for this color, we get an
error, because we didn’t specify a value for yellow.blue:
Download oop/yellow_lightness_fail.cmd
>>> yellow.lightness()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "<stdin>", line 3, in lightness
AttributeError: 'Color' object has no attribute 'blue'
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
CLASS COLOR 275
The solution to all of these problems is to allow users to provide values
for instance variables when new objects are created, like this:
Download oop/color_purple_init.cmd
>>> purple = Color(128, 0, 128)
We have already seen this with classes like str, which can be constructed
with parameters (for example, str(45)). In order to make it work, we need
to add a method to the class Color that will be run whenever a new Color
is created. Such a method is called a constructor, and the way to create
one in Python is to call it __init__:
Download oop/color.py
class Color(object):
'''An RGB color, with red, green and blue components.'''
def __init__(self, r, g, b):
'''A new color with red value r, green value g, and blue value b. All
components are integers in the range 0-255.'''
self.red = r
self.green = g
self.blue = b
The double underscores around the name signal that this method has a
special meaning to Python—in this case, that the method is to be called
when a new object is being created. There are other special methods,
some of which we’ll see later in this chapter. As with other methods,
Python automatically passes a reference to the object itself—in this
case, the one that is being created—as __init__’s first argument. It is then
up to __init__ to set up the object. For example, the expression Color(128,
0, 128) actually means the following:
• Create an object with no instance variables.
• Call Color’s __init__ method with that blank object as the first argu-
ment and the three color values as the second, third, and fourth
arguments.
• Return a reference to that object—which, thanks to __init__, now
has three instance variables—to whoever made the call.
With constructors, methods, and instance variables in hand, we can
now create classes that look and work like those that come with Python
itself.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SPECIAL METHODS 276
13.2 Special Methods
When we introduced __init__ earlier, we said that Python provided other
special methods with underscores around their names as well. We can
use these special methods to make our types look and act more like
Python’s built-in types, which in turn can make our programs easier to
understand.
For example, the output Python produces when we print a Color isn’t
particularly useful:
Download oop/color_purple_no_str.cmd
>>> purple = Color(128, 0, 128)
>>> print purple
<color.Color object at 0x6b150>
This is behavior that object defines for converting objects to strings.
Since object can’t know anything about the meaning of the classes we
will derive from it, it just shows us where the object is in memory. If
we want to present a more useful string, we need to explore two more
special methods, __str__ and __repr__. __str__ is called when an informal,
human-readable version of an object is needed, and __repr__ is called
when more precise, but possibly less-readable, output is desired. In
particular, __str__ is called when print is used.
Let’s define Color.__str__ to provide useful output:
Download oop/color_str.py
class Color(object):
'''An RGB color, with red, green and blue components.'''
def __init__(self, r, g, b):
'''A new color with red value r, green value g, and blue value b. All
components are integers in the range 0-255.'''
self.red = r
self.green = g
self.blue = b
def __str__(self):
'''Return a string representation of this Color in the form of an RGB
tuple.'''
return '(%s, %s, %s)' % (self.red, self.green, self.blue)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SPECIAL METHODS 277
Printing Color now gives more useful information:
Download oop/color_print.cmd
>>> purple = Color(128, 0, 128)
>>> print purple
(128, 0, 128)
Python has lots of other special methods; the official Python website
gives a full list. Among them are __add__, __sub__, and __eq__, which are
called when we add objects with +, subtract them with -, or compare
them with ==:
Download oop/color_full.py
class Color(object):
'''An RGB color, with red, green and blue components.'''
def __init__(self, r, g, b):
'''A new color with red value r, green value g, and blue value b.
All components are integers in the range 0-255.'''
self.red = r
self.green = g
self.blue = b
def __str__(self):
'''Return a string representation of this Color in the form
Color(red, green, blue).'''
return 'Color(%s, %s, %s)' % (self.red, self.green, self.blue)
def __add__(self, other_color):
'''Return a new Color made from adding the red, green, and blue
components of this Color to Color other_color's components. If the
sum is greater than 255, then the color is set to 255.'''
return Color(min(self.red + other_color.red, 255),
min(self.green + other_color.green, 255),
min(self.blue + other_color.blue, 255))
def __sub__(self, other_color):
'''Return a new Color made from subtracting the red, green, and blue
components of this Color from Color other_color's components. If
the difference is less than 0, then the color is set to 0.'''
return Color(max(self.red - other_color.red, 0),
max(self.green - other_color.green, 0),
max(self.blue - other_color.blue, 0))
def __eq__(self, other_color):
'''Return True if this Color's components are equal to Color
other_color's components.'''
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MORE ABOUT DIR AND HELP 278
return self.red == other_color.red and self.green == \
other_color.green and self.blue == other_color.blue
This example shows these methods in action:
Download oop/color_full_use.cmd
>>> purple = Color(128, 0, 128)
>>> white = Color(255, 255, 255)
>>> dark_grey = Color(50, 50, 50)
>>> print purple + dark_grey
Color(178, 50, 178)
>>> print white - dark_grey
Color(205, 205, 205)
>>> print white == Color(255, 255, 255)
True
As you can see, if bright and dark are both colors, then Python inter-
prets the expression bright+dark to mean bright.__add__(dark). This is
called operator overloading, since we are overloading (that is, giving new
meaning to) Python’s built-in operators.
Operator overloading is a powerful tool but should be used very care-
fully. Take addition—we could easily define __add__ so that it modified
the object it was being called on, instead of creating a new one:
Download oop/color_add_bad.py
class Color(object):
...other definitions as before...
def __add__(self, other_color):
'''This is a bad way to define this method.'''
self.red += other_color.red
self.green += other_color.green
self.blue += other_color.blue
return self
Python won’t stop us from doing this—it doesn’t understand the in-
tended meaning, or semantics, of our code. However, just as nobody
would expect adding the integers i and j to change the value of i, nobody
would expect adding colors to change the colors being added. If our
definitions of operators don’t meet people’s expectations, it is our fault,
not theirs, when they use those operators incorrectly.
13.3 More About dir and help
We discussed functions help and dir in Section 4.1, Importing Modules,
on page 50 and the sidebar on page 55. Notice that the contents of a
Color object include the instance variables and all the Color methods.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MORE ABOUT DIR AND HELP 279
Download oop/color_dir.cmd
>>> black = Color(0, 0, 0)
>>> dir(black)
['__add__', '__class__', '__delattr__', '__dict__', '__doc__', '__eq__',
'__getattribute__', '__hash__', '__init__', '__module__', '__new__',
'__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__str__',
'__sub__', '__weakref__', 'blue', 'distance', 'green', 'red']
Since we have written docstrings, we can get help on class Color:
Download oop/color_help.cmd
>>> help(Color)
Help on class Color in module color_full_distance:
class Color(__builtin__.object)
| An RGB color, with red, green and blue components.
|
| Methods defined here:
|
| __add__(self, other_color)
| Return a new Color made from adding the red, green, and blue
| components of this Color to Color other_color's components.
| If the sum is greater than 255, then the color is set to 255.
|
| __eq__(self, other_color)
| Return True if this Color's components are equal to Color
| other_color's components.
|
| __init__(self, r, g, b)
| A new color with red value r, green value g, and blue value b. All
| components are integers in the range 0-255.
|
| __str__(self)
| Return a string representation of this Color in the form
| Color(red, green, blue).
|
| __sub__(self, other_color)
| Return a new Color made from subtracting the red, green, and blue
| components of this Color from Color other_color's components.
| If the difference is less than 0, then the color is set to 0.
|
| lightness(self)
| Return the lightness of this color.
| ----------------------------------------------------------------------
| Data descriptors defined here:
|
| __dict__
| dictionary for instance variables (if defined)
|
| __weakref__
| list of weak references to the object (if defined)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 280
Toward the end of the help text, there is a __dict__ item described, and
it says “dictionary for instance variables.” Python implements instance
variables using dictionaries, and you can inspect this variable:
Download oop/color_dict.cmd
>>> black.__dict__
{'blue': 0, 'green': 0, 'red': 0}
Whenever you assign to an instance variable, it changes the contents
of the object’s dictionary. You can even change it yourself directly,
although we don’t recommend it.
Before looking at another class example, let’s take a short detour and
look at some of the theory behind object-oriented programming.
13.4 A Little Bit of OO Theory
Classes and objects are programming’s power tools. They let good pro-
grammers do a lot in very little time, but with them, bad programmers
can create a real mess. This section will introduce some underlying the-
ory that will help you design reliable, reusable object-oriented software.
Encapsulation
To encapsulate something means to enclose it in some kind of con-
tainer. In programming, encapsulation means keeping data and the
code that uses it in one place and hiding the details of exactly how
they work together. For example, each instance of class file keeps track
of what file on disk it is reading or writing and where it currently is in
that file. The class hides the details of how this is done so that pro-
grammers can use it without needing to know the details of how it was
implemented.
One of the biggest benefits of encapsulation is that it allows program-
mers to change their minds about one part of a program without having
to rewrite other parts. For example, suppose we want to represent rect-
angular sections of images.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 281
Our first attempt might store the XY coordinates of the rectangle’s
bottom-left and upper-right corners, like this:
Download oop/rectangle_corners.py
class Rectangle(object):
'''Represent a rectangular section of an image.'''
def __init__(self, x0, y0, x1, y1):
'''Create a rectangle with non-zero area. (x0,y0) is the
lower left corner, (x1,y1) the upper right corner.'''
self.x0 = x0
self.y0 = y0
self.x1 = x1
self.y1 = y1
def area(self):
'''Return the area of the rectangle.'''
return (self.x1 - self.x0) * (self.y1 - self.y0)
def contains(self, x, y):
'''Return True is (x,y) point is inside a rectangle,
and False otherwise.'''
return (self.x0 <= x <= self.x1) and \
(self.y0 <= y <= self.y1)
Later, we might decide that it would be better to store the rectangle’s
lower-left corner and XY size, like this:
Download oop/rectangle_size.py
class Rectangle(object):
'''Represent a rectangular section of an image.'''
def __init__(self, x0, y0, width, height):
'''Create a rectangle with non-zero area. (x0,y0) is the
lower left corner, width and height the X and Y extent.'''
self.x0 = x0
self.y0 = y0
self.width = width
self.height = height
def area(self):
'''Return the area of the rectangle.'''
return self.width * self.height
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 282
def contains(self, x, y):
'''Return True if (x,y) point is inside a rectangle,
and False otherwise.'''
return (self.x0 <= x) and (x <= self.x0 + width) and \
(self.y0 <= y) and (y <= self.y0 + height)
If we made this change, we would obviously also have to change every
piece of software that created new rectangles. However, we wouldn’t
have to change code that used the area or contains methods—since they
hide the details of how they calculate their results, we can change their
operation without affecting anything else.
Polymorphism
Polymorphism means “having more than one form.” In programming, it
means that an expression involving a variable can do different things
depending on the type of the object to which the variable refers. For
example, if obj refers to a string, then obj[1:3] produces a two-character
string. If obj refers to a list, on the other hand, the same expression pro-
duces a two-element list. Similarly, the operator left + right can produce
a number, a string, or a list, depending on the types of left and right.
Polymorphism is used throughout modern programs to cut down on
the amount of code programmers need to write and test. It lets us write
a generic function to count nonblank lines:
Download oop/line_counter.py
def non_blank_lines(thing):
'''Return the number of non-blank lines in thing.'''
count = 0
for line in thing:
if line.strip():
count += 1
return count
and then apply it to a list of strings, a file, a web page on a site halfway
around the world (see Section 8.1, Files Over the Internet, on page 160),
or a single string wrapped up in the StringIO class to look like a file.
Each of those four types knows how to be the subject of a loop; in other
words, each one knows how to produce its “next” element as long as
there is one and then say “all done.” That means that instead of writing
four functions to count interesting lines or copying the lines into a list
and then applying one function to that list, we can apply one function
to all those types directly.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 283
Languages like Java and C++ require programmers to declare in ad-
vance that classes are polymorphic to one another (usually by using
inheritance, which is described in a moment). Python has a more re-
laxed approach. If the classes provide methods with the same names
and the same number of arguments, then instances of one can be sub-
stituted for instances of the other without any extra work. This means
that if you want to create something that can be used in place of a file,
all you have to do is give it methods like read and readlines. Similarly, if
someone else wants to create their own kind of rectangles for your code
to use, all they have to do is provide the right kind of constructor and
the methods area and contains.
At least, that’s all they should have to do. Right now, though, the class
Rectangle isn’t properly encapsulated, so it would actually be harder
for someone to write a replacement than it ought to be. The reason
is that right now, users of the class have to deal with the rectangle’s
implementation to get its corner points. In our first implementation,
for example, a user would write r.x1 to get the rectangle’s maximum X
extent, while in the second case, she would write r.x0 + r.width.
The way to make polymorphism easier is to encapsulate all of the imple-
mentation details by only ever talking to class instances through meth-
ods. If we do this, our first implementation becomes as follows:
Download oop/rectangle_corners_enc.py
class Rectangle(object):
'''Represent a rectangular section of an image.'''
def __init__(self, x0, y0, x1, y1):
...as before...
def area(self):
...as before...
def contains(self, x, y):
...as before...
def get_min_x(self):
'''Return the minimum X coordinate.'''
return self.x0
def get_min_y(self):
'''Return the minimum Y coordinate.'''
return self.y0
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 284
def get_max_x(self):
'''Return the maximum X coordinate.'''
return self.x1
def get_max_y(self):
'''Return the maximum Y coordinate.'''
return self.y1
while our second becomes the following:
Download oop/rectangle_size_enc.py
class Rectangle(object):
'''Represent a rectangular section of an image.'''
def __init__(self, x0, y0, width, height):
...as before...
def area(self):
...as before...
def contains(self, x, y):
...as before...
def get_min_x(self):
'''Return the minimum X coordinate.'''
return self.x0
def get_min_y(self):
'''Return the minimum Y coordinate.'''
return self.y0
def get_max_x(self):
'''Return the maximum X coordinate.'''
return self.x0 + width
def get_max_y(self):
'''Return the maximum Y coordinate.'''
return self.y0 + height
More advanced programmers usually take advantage of a built-in func-
tion in Python called property to do some of the work that we’re doing
by hand in these examples; see the Python library documentation for
details and examples.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 285
Inheritance
Giving one class the same methods as another is one way to make them
polymorphic, but it suffers from the same flaw as initializing an object’s
instance variables from outside the object. If a programmer forgets just
one line of code, the whole program can fail for reasons that will be
difficult to track down. A better approach is to use a third fundamental
feature of object-oriented programming called inheritance, which allows
you to recycle code in yet another way.
To explore how this is done, suppose we are trying to simulate the
ecosystem of a tide pool. The pool contains all sorts of living things
that float around and interact with their neighbors. Each critter moves
around in its own special way and has its own favorite food. We define
a class Organism that represents this living thing:
Download oop/organism.py
class Organism(object):
'''A thing that lives in a tide pool.'''
def __init__(self, name, x, y):
'''A living thing that is at location (x,y) in the tide pool.'''
self.name = name
self.x = x
self.y = y
def __str__(self):
'''Return a string representation of this Organism.'''
return '(%s, [%s, %s])' % (self.name, self.x, self.y)
def can_eat(self, food):
'''Report whether this Organism can eat the given food.
Since we don't know anything about what a generic organism
eats, this always returns False.'''
return False
def move(self):
'''Ask the organism to move. By default, this does nothing,
since we don't know anything about how fast or how far a
generic organism would move.'''
return
The class Organism has three instance variables: name, x, and y (the
latter two representing its coordinates within the pool). It also has four
methods: __init__, __str__, can_eat, and move. Since we don’t know any-
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 286
thing about the Organism, it never actually moves and doesn’t eat any-
thing.
Not all real organisms behave like our generic Organism. Crabs walk,
green algae float, and mussels swim—though not the same way fish do.
Some eat plants; some eat animals; green algae “eat” sunlight through
photosynthesis.
We can define these specific types of organisms as new classes. Rather
than defining each one from scratch, we start with our generic organism
by replacing object with Organism in the class header:
Download oop/arthropod_header.py
class Arthropod(Organism):
pass
Here, the keyword pass means “do nothing to the definitions inherited
from Organism.” We call Organism the parent or superclass, and we call
Arthropod the child or subclass. Arthropod inherits instance variables
and methods from its parent Organism, so an Arthropod object automati-
cally has the instance variables name, x, and y and the methods __init__,
__str__, can_eat, and move:
Download oop/arthropod_basic.py
>>> blue_crab = Arthropod('Callinectes sapidus', 0, 0)
>>> print blue_crab
(Callinectes sapidus, 0, 0)
However, we want the Arthropod class to be more than just a generic
Organism, which means giving it its own instance variables, methods,
or both. In this case, a leg_count instance variable is needed, which
requires changing the definition of __init__:
Download oop/arthropod_init.py
class Arthropod(Organism):
'''An arthropod that has a fixed number of legs.'''
def __init__(self, name, x, y, legs):
'''An arthropod with the given number of legs that exists at location
(x, y) in the tide pool.'''
Organism.__init__(self, name, x, y)
self.legs = legs
In the first line of __init__, we call the constructor from class Organ-
ism to initialize the three instance variables, which are name, x, and y.
The second line creates the leg_count instance variable and initializes
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LITTLE BIT OF OO THEORY 287
it. Arthropod.__init__ overrides Organism __init__, so when an instance of
Arthropod calls __init__, the version defined in that class is called, rather
than the one in the parent class:
Download oop/arthropod_init_result.py
>>> lobster = Arthropod('Homarus gammarus', 0, 0)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: __init__() takes exactly 5 arguments (4 given)
>>> lobster = Arthropod('Homarus gammarus', 0, 0, 10)
The call to the Arthropod constructor with only three arguments results
in an error. Once a function is overridden, the parent’s version of the
function cannot be called from outside the class definition.
Notice, though, that when we print an instance of Arthropod, the num-
ber of legs isn’t reported:
Download oop/arthropod_str.py
>>> lobster = Arthropod('Homarus gammarus', 0, 0, 10)
>>> print lobster
(Homarus gammarus, 0, 0)
This happens because we haven’t overridden the __str__ method, so
Organism.__str__ is still being called. Fixing this is straightforward:
Download oop/arthropod_str_override.py
class Arthropod(Organism):
'''An arthropod that has a fixed number of legs.'''
def __init__(self, name, x, y, legs):
'''An arthropod with the given number of legs that exists at location
(x, y) in the tide pool.'''
Organism.__init__(self, name, x, y)
self.legs = legs
def __str__(self):
'''Return a string representation of this Arthropod.'''
return '(%s, %s, [%s, %s])' % (self.name, self.legs, self.x, self.y)
Of course, a child class can also have methods that are not part of the
parent class:
Download oop/arthropod.py
class Arthropod(Organism):
'''An arthropod that has a fixed number of legs.'''
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LONGER EXAMPLE 288
Figure 13.2: Jmol: an open source Java viewer for chemical structures
in 3D
def __init__(self, name, x, y, legs):
'''An arthropod with the given number of legs that exists at location
(x, y) in the tide pool.'''
Organism.__init__(self, name, x, y)
self.legs = legs
def __str__(self):
'''Return a string representation of this Arthropod.'''
return '(%s, %s, [%s, %s])' % (self.name, self.legs, self.x, self.y)
def is_decapod(self):
'''Return True if this Arthropod is a decapod.'''
return self.legs == 10
def leg_count(self):
'''Return the number of legs this Arthropod possesses.'''
return self.legs
13.5 A Longer Example
Molecular graphic visualization tools allow for interactive exploration
of molecular structures. Most read PDB-formatted files, which we de-
scribe in Section 8.4, Multiline Records, on page 177.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LONGER EXAMPLE 289
In a molecular visualizer, every atom, molecule, bond, and so on, has
a location in 3D space, usually defined as a vector, which is an arrow
from the origin to where the structure is. All of these structures can be
rotated and translated.
A vector is usually represented by x, y, and z coordinates that specify
how far along the x-axis, y-axis, and z-axis the vector extends.
Here is how ammonia can be specified in PDB format:
Download fileproc/ammonia.pdb
COMPND AMMONIA
ATOM 1 N 0.257 -0.363 0.000
ATOM 2 H 0.257 0.727 0.000
ATOM 3 H 0.771 -0.727 0.890
ATOM 4 H 0.771 -0.727 -0.890
END
In our simplified PDB format, a molecule is made up of numbered
atoms. In addition to the number, an atom has a symbol and (x, y, z)
coordinates. For example, in one of the atoms in ammonia is nitrogen
with symbol N at coordinates (0.257, -0.363, 0.0). In the following sections,
we will look at how we could translate these ideas into object-oriented
Python.
Class Atom
We might want to create an atom like this, using information we read
from the PDB file:
Download oop/atom.py
nitrogen = Atom(1, "N", 0.257, -0.363, 0.0)
To do this, we’ll need a class called Atom with a constructor that creates
all the appropriate instance variables:
Download oop/atom.py
class Atom(object):
'''An atom with a number, symbol, and coordinates.'''
def __init__(self, num, sym, x, y, z):
'''Create an Atom with number num, string symbol sym, and float
coordinates (x, y, z).'''
self.number = num
self.center = (x, y, z)
self.symbol = sym
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LONGER EXAMPLE 290
To inspect an Atom, we’ll want to provide __repr__ and __str__ methods:
Download oop/atom.py
def __str__(self):
'''Return a string representation of this Atom in this format:
(SYMBOL, X, Y, Z)
'''
return '(%s, %s, %s, %s)' % \
(self.symbol, self.center[0], self.center[1], self.center[2])
def __repr__(self):
'''Return a string representation of this Atom in this format:
Atom("SYMBOL", X, Y, Z)
'''
return 'Atom(%s, "%s", %s, %s, %s)' % \
(self.number, self.symbol, \
self.center[0], self.center[1], self.center[2])
We’ll use those later, when we define a class for molecules.
In visualizers, one common operation is translation: move an atom to
a different location. We’d like to be able to write this in order to tell the
nitrogen atom to move up by 0.2 units:
Download oop/atom.py
nitrogen.translate(0, 0, 0.2)
This code works as expected if we add the following method to class
Atom:
Download oop/atom.py
def translate(self, x, y, z):
'''Move this Atom by adding (x, y, z) to its coordinates.'''
self.center = (self.center[0] + x,
self.center[1] + y,
self.center[2] + z)
Class Molecule
Remember that we read PDB files one line at a time. When we reach
the line containing COMPND AMMONIA, we know that we’re building a
complex structure: a molecule with a name and a list of atoms.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LONGER EXAMPLE 291
Here’s the start of a class for this, including an add method that adds
an Atom to the molecule:
Download oop/molecule.py
class Molecule(object):
'''A molecule with a name and a list of Atoms.'''
def __init__(self, name):
'''Create a Compound named name with no Atoms.'''
self.name = name
self.atoms = []
def add(self, a):
'''Add Atom a to my list of Atoms.'''
self.atoms.append(a)
As we read through the ammonia PDB information, we add atoms as
we find them; here is the code from Section 8.4, Multiline Records, on
page 177 rewritten to return a Molecule object instead of a list of tuples:
Download oop/multimol_2.py
from molecule import Molecule
from atom import Atom
def read_molecule(r):
'''Read a single molecule from reader r and return it,
or return None to signal end of file.'''
# If there isn't another line, we're at the end of the file.
line = r.readline()
if not line:
return None
# Name of the molecule: "COMPND name"
key, name = line.split()
# Other lines are either "END" or "ATOM num kind x y z"
molecule = Molecule(name)
reading = True
while reading:
line = r.readline()
if line.startswith('END'):
reading = False
else:
key, num, kind, x, y, z = line.split()
molecule.add(Atom(num, kind, float(x), float(y), float(z)))
return molecule
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A LONGER EXAMPLE 292
If we compare the two versions, we can see the code is nearly identical.
It’s just as easy to read the new version as the old—more so even,
because it includes type information. Here are the __str__ and __repr__
methods:
Download oop/molecule.py
def __repr__(self):
'''Return a string representation of this Molecule in this format:
Molecule("NAME", (ATOM1, ATOM2, ...))
'''
res = ''
for atom in self.atoms:
res = res + repr(atom) + ', '
# Strip off the last comma.
res = res[:-2]
return 'Molecule("%s", (%s))' % (self.name, res)
def __str__(self):
'''Return a string representation of this Molecule in this format:
(NAME, (ATOM1, ATOM2, ...))
'''
res = ''
for atom in self.atoms:
res = res + str(atom) + ', '
# Strip off the last comma.
res = res[:-2]
return '(%s, (%s))' % (self.name, res)
We’ll add a translate method to Molecule to make it easier to move:
Download oop/molecule.py
def translate(self, x, y, z):
'''Move this Compound, including all Atoms, by (x, y, z).'''
for atom in self.atoms:
atom.translate(x, y, z)
And here we call it:
Download oop/molecule.py
ammonia = Molecule("AMMONIA")
ammonia.add(Atom(1, "N", 0.257, -0.363, 0.0))
ammonia.add(Atom(2, "H", 0.257, 0.727, 0.0))
ammonia.add(Atom(3, "H", 0.771, -0.727, 0.890))
ammonia.add(Atom(4, "H", 0.771, -0.727, -0.890))
ammonia.translate(0, 0, 0.2)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 293
13.6 Summary
In this chapter, we learned the following:
• In object-oriented languages, new types are defined by creating
classes. Classes support encapsulation; in other words, they com-
bine data and the operations on it so that other parts of the pro-
gram can ignore implementation details.
• Classes also support polymorphism. If two classes have methods
that work the same way, instances of those classes can replace
one another without the rest of the program being affected. This
enables “plug-and-play” programming, in which one piece of code
can perform different operations depending on the objects it is
operating on.
• Finally, new classes can be defined by inheriting features from
existing ones. The new class can override the features of its parent
and/or add entirely new features.
• When a method is defined in a class, its first argument must be a
variable that represents the object the method is being called on.
By convention, this argument is called self.
• Some methods have special predefined meanings in Python; to sig-
nal this, their names begin and end with two underscores. Some
of these methods are called when constructing objects (__init__) or
converting them to strings (__str__ and __repr__); others, like __add__
and __sub__, are used to imitate arithmetic.
13.7 Exercises
The best way to learn how to do object-oriented programming is to
go back through the examples and exercises of previous chapters and
see which ones are easier or more naturally written using classes and
objects. You can also create object-oriented programs that represent
things in the real world. For example, what classes would you use to
model the stars, planets, moons, rings, and comets that make up a
solar system? What methods should each class have? How and where
would you keep track of things like orbital parameters? How would you
prevent programmers from accidentally putting a star in orbit around
a comet or something equally silly?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 14
Graphical User InterfacesMost of the programs in previous chapters are not interactive. Once
launched, they run to completion without giving us a chance to steer
them or provide new input. The few that do communicate with us do so
through the kind of text-only command-line user interface, or CLUI, that
would have already been considered old-fashioned in the early 1980s.
As you already know, most modern programs interact with users via a
graphical user interface, or GUI, which is made up of windows, menus,
buttons, and so on. In this chapter, we will show you how to build sim-
ple GUIs using a Python module called Tkinter. Along the way, we will
introduce a different way of structuring programs called event-driven
programming. A traditionally structured program usually has control
over what happens when, but an event-driven program must be able to
respond to input at unpredictable moments. As we shall see, the easiest
way to do this is to use some of Python’s more advanced features.
Tkinter is one of several toolkits you can use to build GUIs in Python,
and other languages have toolkits of their own. However, knowing how
to put buttons and sliders on the screen is only part of knowing how to
create an application that people will understand, use, and enjoy. To do
the latter, you will also need to know something about graphic design
and other aspects of human-computer interaction. Hundreds of books
have been written on the subject (we particularly like [Joh07]), and you
can find hundreds of tutorials and resources on the Web.
Prepared exclusively for Trieu Nguyen
THE TKINTER MODULE 295
Figure 14.1: A root window
14.1 The Tkinter Module
Tkinter comes with Python and allows us to create windows, buttons,
menus, text areas, checkboxes, and many other widgets. We usually
start by importing everything in the Tkinter module into our program:1
Download gui/import.py
from Tkinter import *
Every Tkinter GUI application starts by creating a root window and
saving a reference to it in a variable:
Download gui/getwindow.py
window = Tk()
The root window is initially empty (see Figure 14.1). It acts as the out-
ermost container for the application; all the other widgets we create will
depend on it. If the window on the screen is closed, the window object
is destroyed (though we can create a new root window by calling Tk()
again). All of the applications we will create have only one main win-
dow, but additional windows can be created using the TopLevel widget
(see Figure 14.2, on the next page).
1. We said in Section 4.1, Importing Modules, on page 50 that import * was bad style
because of the risk of name collision. We are using it here because most online tutorials
about Tkinter are written this way. In your applications, you should use import Tkinter as tk
and then use tk.thing to refer to its elements.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BASIC GUI CONSTRUCTION 296
Widget Description
Button A clickable button
Canvas An area used for drawing or displaying images
Checkbutton A clickable box that can be selected or unselected
Entry A single-line text field that the user can type in
Frame A container for widgets
Label A single-line display for text
Listbox A drop-down list that the user can select from
Menu A drop-down menu
Message A multiline display for text
Menubutton An item in a drop-down menu
Text A multiline text field that the user can type in
TopLevel An additional window
Figure 14.2: Tkinter widgets
14.2 Basic GUI Construction
The next step in developing a GUI is to add widgets to the root window.
The simplest widget is a Label, which is used to display short pieces of
text. When we create one, we must specify its parent widget, which is
the widget that the label is placed inside. In our case, the parent widget
is the root window of the application. We also need to provide the text
that is to be displayed, which can be done by setting the label’s text
attribute when we construct it. (By convention, we do not place spaces
on either side of the equal sign when setting the value of attributes like
text.)
Download gui/label.py
from Tkinter import *window = Tk()
label = Label(window, text="This is our label.")
label.pack()
The last line of this little program is crucial. Like other widgets, Label
has a method called pack that places it in its parent and then tells
the parent to resize itself as necessary. If we forget to call this method,
the child widget (in this case, the label) won’t be displayed or will be
displayed improperly. After label.pack() is called, on the other hand, we
see the GUI shown in Figure 14.3, on the following page.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BASIC GUI CONSTRUCTION 297
Figure 14.3: A window with a label
Labels are often used to display static text that never changes, such as
copyright information. Often, though, applications will want to update
a label’s text as the program runs to show things like the currently
open file or the time of day. One way to do this is simply to assign a
new value to the widget’s text using named parameters (Section 12.1,
Named Parameter, on page 241):
Download gui/label-dict.py
from Tkinter import *import time
window = Tk()
label = Label(window, text="First label.")
label.pack()
time.sleep(2)
label.config(text="Second label.")
Run the previous code one line at a time from the Python command
prompt to see how the label changes. This code will not display the
window at all if run all at once as a script or from within Wing 101. In
those cases, the call to window.mainloop() is needed to tell the program
to pay attention to the outside world.
Mutable Variables
There is a better way to manage the interactions between a program’s
GUI and its variables, and the reason it’s better brings us face to face
with the biggest difference between the applications we have seen in
previous chapters and event-driven GUI applications. Suppose we want
to display a string, such as the current time, in several places in a
GUI—the application’s status bar, some dialog boxes, and so on. As-
signing a new value to each widget each time the string changes isn’t
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BASIC GUI CONSTRUCTION 298
Immutable Type Tkinter Mutable Type
int IntVar
string StringVar
bool BooleanVar
double DoubleVar
Figure 14.4: Tkinter mutable types
hard, but as the application grows, so too do the odds that we’ll forget
to update at least one of the widgets that’s displaying the string. What
we really want is a string that “knows” which widgets care about its
value and can update them itself when that value changes.
Since Python’s strings, integers, doubles, and Booleans are immutable,
Tkinter provides types of its own that can be updated in place and
that can notify widgets whenever their values change (see Figure 14.4).
Rather than set the text of the label using an immutable type such
as string, we can set it using the corresponding mutable type, such as
StringVar. Whenever a new value is assigned to that StringVar, it tells the
label, and any other widgets it has been assigned to, that it’s time to
update.
The values in Tkinter mutable types are set and retrieved using the set
and get methods. To show how they work, the following code creates a
Tkinter string variable called data and sets its value to "Data to display".
It then creates a label to display the contents of data:
Download gui/label-variable.py
from Tkinter import *window = Tk()
data = StringVar()
data.set("Data to display")
label = Label(window, textvariable=data)
label.pack()
window.mainloop()
Notice that this time we set the textvariable attribute of the label rather
than the text attribute. Any time the program changes the contents of
data, the text the label is displaying will automatically change as well.
The relationships between the three main variables in this program are
shown in Figure 14.5, on the next page.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BASIC GUI CONSTRUCTION 299
data
label
'Data to display'
window
Figure 14.5: Widgets and mutable variables
There is one small trap here for newcomers: because of the way the
Tkinter module is structured, you cannot create a StringVar or any other
mutable variable until you have called the Tk() function to create the
top-level window. This doesn’t make much difference in this case, but
as we’ll see in a moment, it sometimes forces programmers to do things
in an unintuitive order.
Frame
To show the real power of mutable variables, we need to create a GUI
that has several widgets. We will use this as an opportunity to introduce
another widget called Frame. A frame isn’t directly visible on the screen;
instead, its purpose is to organize other widgets. To create a GUI that
displays three labels, for example, the following code puts a frame in
the root window and then adds the labels to the frame one by one:
Download gui/frame.py
from Tkinter import *
window = Tk()
frame = Frame(window)
frame.pack()
first = Label(frame, text="First label")
first.pack()
second = Label(frame, text="Second label")
second.pack()
third = Label(frame, text="Third label")
third.pack()
window.mainloop()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
BASIC GUI CONSTRUCTION 300
Figure 14.6: A window with a frame and three labels
The resulting GUI is shown in Figure 14.6, while in Figure 14.7, on the
next page, we can see how the five widgets are organized. Note that we
call pack on every widget; if we omit one of these calls, that widget will
not be displayed.
Putting the three labels in the same frame is equivalent to putting the
labels directly into the main window widget, but we can use multiple
frames to format the window’s layout. Here we use two frames con-
taining the three labels and put a border around the second frame. We
specify the border width using the borderwidth attribute (0 is the default)
and the border style using relief (FLAT is the default). The other border
styles are SUNKEN, RAISED, GROOVE, and RIDGE.
Download gui/frame2.py
window = Tk()
frame = Frame(window)
frame.pack()
frame2 = Frame(window, borderwidth=4, relief=GROOVE)
frame2.pack()
first = Label(frame, text="First label")
first.pack()
second = Label(frame2, text="Second label")
second.pack()
third = Label(frame2, text="Third label")
third.pack()
window.mainloop()
Entry
Two widgets let users enter text. The simplest one is Entry, which allows
for a single line of text. If we associate a StringVar with the Entry, then
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODELS, VIEWS, AND CONTROLLERS 301
window
frame
first
second
third
Figure 14.7: Structure of a three-label GUI
whenever a user types anything into that Entry, the StringVar’s value will
automatically be updated to the contents of the Entry. If that same String-
Var is also bound to a label, then the label will display whatever is cur-
rently in the Entry without us doing any extra work. The following code
shows how to set this up, and we can see what the GUI looks like in
Figure 14.8, on the following page:
Download gui/entry.py
from Tkinter import *
window = Tk()
frame = Frame(window)
frame.pack()
var = StringVar()
label = Label(frame, textvariable=var)
label.pack()
entry = Entry(frame, textvariable=var)
entry.pack()
window.mainloop()
14.3 Models, Views, and Controllers
Using a StringVar to connect a text-entry box and a label is the first
step toward separating models, views, and controllers, which is the key
to building larger GUIs (and many other kinds of applications). As its
name suggests, a view is something that displays information to the
user, like Label.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODELS, VIEWS, AND CONTROLLERS 302
Figure 14.8: An entry and a label tied together
Many views, like Entry, also accept input, which they display immedi-
ately. The key is that they don’t do anything else: they don’t calculate
average temperatures, move robot arms, or do any other calculations.
Models, on the other hand, just store data, like a piece of text or the
current inclination of a telescope. They also don’t do calculations; their
job is simply to keep track of the application’s current state (and, in
some cases, to save that state to a file or database and reload it later).
Sitting beside an application’s models and views are its controllers,
implementing its intelligence. The controller is what decides whether
two gene sequences match well enough to be colored green or whether
someone is allowed to overwrite an old results file. Controllers may
update an application’s models, which in turn can trigger changes to
its views.
The following code shows what all of this looks like in practice. Here,
the model is a simple integer counter, which is implemented as an IntVar
so that the view will update itself automatically. The controller is the
function click, which updates the model whenever a button is clicked.
Four objects make up the view: the root window, a Frame, a Label that
shows the current value of counter, and a button that the user can click
to increment counter’s value:
Download gui/mvc.py
# Initialization.
from Tkinter import *
# The controller.
def click():
counter.set(counter.get() + 1)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODELS, VIEWS, AND CONTROLLERS 303
if __name__ == '__main__':
# More initialization.
window = Tk()
# The model.
counter = IntVar()
counter.set(0)
# The views.
frame = Frame(window)
frame.pack()
button = Button(frame, text="Click", command=click)
button.pack()
label = Label(frame, textvariable=counter)
label.pack()
window.mainloop()
The first two arguments used to construct the Button should be familiar
by now. The third, command=click, tells it to call the function click each
time the user presses the button. This makes use of the fact that, in
Python, a function is just another kind of object and can be passed as
an argument like anything else.
The click function in the previous code does not have any parame-
ters but uses the variable counter that is defined outside the function.
Variables like this are called global variables, and their use should be
avoided, since they make programs hard to understand. It would be
better to pass any variables the function needs into it as parameters.
We can’t do this using the tools we have seen so far, because the func-
tions that our buttons can call must not have any parameters. We will
show you one way to avoid using global variables in the next section,
and we’ll show you another in Section 13.1, Class Color, on page 271.
Using Lambda
The simple counter GUI shown earlier does what it’s supposed to, but
there is room for improvement. For example, suppose we want to be
able to lower the counter’s value as well as raise it.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODELS, VIEWS, AND CONTROLLERS 304
Using only the tools we have seen so far, we could add another button
and another controller function like this:
Download gui/mvc2.py
# Initialization.
from Tkinter import *window = Tk()
# The model.
counter = IntVar()
counter.set(0)
# Two controllers
def click_up():
counter.set(counter.get() + 1)
def click_down():
counter.set(counter.get() - 1)
# The views.
frame = Frame(window)
frame.pack()
button = Button(frame, text="Up", command=click_up)
button.pack()
button = Button(frame, text="Down", command=click_down)
button.pack()
label = Label(frame, textvariable=counter)
label.pack()
window.mainloop()
This seems a little clumsy, though. The two functions click_up and
click_down are doing almost the same thing; surely we ought to be able
to combine them into one. While we’re at it, we ought to pass counter
into the function explicitly, rather than using it as a global variable:
Download gui/mvc_one_func.py
# The model.
counter = IntVar()
counter.set(0)
# One controller with parameters.
def click(variable, value):
variable.set(variable.get() + value)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODELS, VIEWS, AND CONTROLLERS 305
The problem with this is figuring out what to pass into the buttons,
since we can’t provide any arguments for the functions assigned to
the buttons’ command attributes when creating those buttons.2 Tkinter
cannot read our minds—it can’t magically know how many arguments
our functions require or what values to pass in for them. For that rea-
son, it requires that the controller functions triggered by buttons and
other widgets take zero arguments so they can all be called the same
way. It is our job to figure out how to take the two-argument function
we want to use and turn it into one that needs no arguments at all.
We could do this by writing a couple of wrapper functions:
Download gui/mvc_wrapper_func.py
def click_up():
click(counter, 1)
def click_down():
click(counter, -1)
but this gets us back to two nearly identical functions that rely on
global variables. A better way is to use a lambda function, which allows
us to create a one-line function anywhere we want without giving it a
name.3 Here’s a very simple example:
Download gui/lambda_1.py
>>> lambda: 3
<function <lambda> at 0x00A89B30>
>>> (lambda: 3)()
3
The expression lambda: 3 on the first line creates a nameless function
that always returns the number 3. The second expression creates this
function and immediately calls it, which has the same effect as this:
Download gui/lambda_2.py
>>> def f():
... return 3
...
>>> f()
3
2. If we tried to pass parameters to click_up when creating the button, then Python would
call click_up and assign the result of the call to the command parameter, instead of assign-
ing the function itself to command.
3. The name lambda function comes from the Lambda Calculus, a mathematical system
for investigating function definition and application that was developed in the 1930s by
Alonzo Church and Stephen Kleene.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
MODELS, VIEWS, AND CONTROLLERS 306
However, the lambda form does not create a new variable or change an
existing one. Finally, lambda functions can take arguments, just like
other functions:
Download gui/lambda_4.py
>>> (lambda x: 2 * x)(3)
6
So, how does this help us with GUIs? The answer is that it lets us write
one controller function to handle different buttons in a general way
and then wrap up calls to that function when and as needed. Here’s
the two-button GUI once again using lambda functions:
Download gui/mvc3.py
# Initialization.
from Tkinter import *window = Tk()
# The model.
counter = IntVar()
counter.set(0)
# General controller
def click(var, value):
var.set(var.get() + value)
# The views.
frame = Frame(window)
frame.pack()
button = Button(frame, text="Up", command=lambda: click(counter, 1))
button.pack()
button = Button(frame, text="Down", command=lambda: click(counter, -1))
button.pack()
label = Label(frame, textvariable=counter)
label.pack()
window.mainloop()
This code creates a zero-argument lambda function to pass into each
button just where it’s needed. Those lambda functions then pass the
right values into click. This is cleaner than the preceding code because
the function definitions are enclosed in the call that uses them—there
is no need to clutter the GUI with little functions that are used only in
one place.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE 307
Note, however, that it is a very bad idea to repeat the same function
several times in different places—if you do that, the odds are very high
that you will one day want to change them all but will miss one or two.
If you find yourself wanting to do this, reorganize the code so that the
function is defined only once.
14.4 Style
Every windowing system has its own look and feel—square or rounded
corners, particular colors, and so on. In this section, we will see how to
change the appearance of a GUI’s widgets to make an application look
more distinctive.
A note of caution before we begin: the default styles of most windowing
systems have been chosen by experts trained in graphic design and
human-computer interaction. The odds are that any radical changes on
your part will make things worse, not better. In particular, be careful
about colors (several percent of the male population has some degree
of color blindness) and font sizes (many people, particularly the elderly,
cannot read small text).
Fonts
Let’s start by changing the size, weight, slant, and family of the font
used to display text. To specify the size, we provide the height as an
integer in points. We can set the weight to either bold or normal and
the slant to either italic (slanted) or roman (not slanted).
The font families we can use depend on what system the program is
running on. Common families include Times, Courier, and Verdana,
but dozens of others are usually available. One note of caution, though:
if you choose an unusual font, people running your program on other
computers might not have it, so your GUI might appear different than
you’d like for them.
The following sets the font of a button to be 14-point, bold, italic, and
Courier. The result is shown in Figure 14.9, on the following page.
Download gui/font.py
from Tkinter import *window = Tk()
button = Button(window, text="Hello", font=("Courier", 14, "bold italic"))
button.pack()
window.mainloop()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE 308
Figure 14.9: A button with customized text font
Colors
Almost all widgets have background and foreground colors, which can
be set using the bg and fg attributes, respectively. As the following code
shows, we can set either of these to a standard color by specifying the
color’s name, such as white, black, red, green, blue, cyan, yellow, or
magenta:
Download gui/color.py
from Tkinter import *window = Tk()
button = Label(window, text="Hello", bg="green", fg="white")
button.pack()
window.mainloop()
The result is shown in Figure 14.10, on the next page. As you can
see, pure white text on a bright green background is not particularly
readable.
We can choose more colors by specifying them using the RGB color
model introduced in Section 4.4, Pixels and Colors, on page 68. As we
said there, RGB values are conventionally written in hexadecimal (base-
16) notation; the best way to understand them is to play with them.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE 309
Figure 14.10: A label with color
The following color picker does this by updating a piece of text to show
the color specified by the red, green, and blue values entered in the text
boxes:
Download gui/colorpicker.py
from Tkinter import *
def change(widget, colors):
'''Update the foreground color of a widget to show the RGB color value
stored in a dictionary with keys 'red', 'green', and 'blue'. Does
*not* check the color value.
'''
new_val = '#'
for name in ('red', 'green', 'blue'):
new_val += colors[name].get()
widget['bg'] = new_val
# Create the application.
window = Tk()
frame = Frame(window)
frame.pack()
# Set up text entry widgets for red, green, and blue, storing the
# associated variables in a dictionary for later use.
colors = {}
for (name, col) in (('red', '#FF0000'),
('green', '#00FF00'),
('blue', '#0000FF')):
colors[name] = StringVar()
colors[name].set('00')
entry = Entry(frame, textvariable=colors[name], bg=col, fg="white")
entry.pack()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE 310
# Display the current color.
current = Label(frame, text=' ', bg='#FFFFFF')
current.pack()
# Give the user a way to trigger a color update.
update = Button(frame, text='Update', command=lambda: change(current, colors))
update.pack()
# Run the application.
mainloop()
This is the most complicated GUI we have seen so far but can be under-
stood by breaking it down into a model, some views, and a controller.
The model is three StringVars that store the hexadecimal strings repre-
senting the current red, green, and blue components of the color to
display. These three variables are kept in a dictionary indexed by name
for easy access. The controller is the function change, which concate-
nates the strings to create an RGB color and applies that color to the
background of a widget. The views are the text-entry boxes for the color
components, the label that displays the current color, and the button
that tells the GUI to update itself.
This program works, but neither the GUI nor the code is very attractive.
It’s annoying to have to click the update button, and if a user ever types
anything that isn’t a two-digit hexadecimal number into one of the text
boxes, it results in an error. The exercises will ask you to redesign both
the appearance and the structure of this program.
Layout
One of the things that makes the color picker GUI ugly is the fact
that everything is arranged top to bottom. Tkinter uses this layout by
default, but we can usually come up with something better.
To see how, let’s revisit the example from Figure 14.8, on page 302,
placing the label and button horizontally. We tell Tkinter to do this
by providing a side argument to the pack method. The code to do this is
shown here, and the result can be seen in Figure 14.11, on the following
page:
Download gui/side.py
from Tkinter import *window = Tk()
frame = Frame(window)
frame.pack()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
STYLE 311
Figure 14.11: A window with horizontal layout
label = Label(frame, text="Name")
label.pack(side="left")
entry = Entry(frame)
entry.pack(side="left")
window.mainloop()
Setting side to "left" tells Tkinter that the leftmost part of the label is to
be placed next to the left edge of the frame, and then the leftmost part
of the entry field is placed next to the right edge of the label—in short,
that widgets are to be packed using their left edges. We could equally
well pack according to the right, top, or bottom edges, or we could mix
packings (though that can quickly become confusing).
For even more control of our window layout, we can use a different
layout manager called grid. As its name implies, it treats windows and
frames as grids of rows and columns. To add the widget to the window,
we call grid instead of pack. Do not call both! The grid call can take
several parameters, which are shown in Figure 14.12, on the next page.
In the following code, we place the label in the upper left (row 0, column
0) and the entry field in the lower right (row 1, column 1). As you can
see in Figure 14.13, on page 313, this leaves the bottom-left and upper-
right corners empty:
Download gui/grid.py
from Tkinter import *window = Tk()
frame = Frame(window)
frame.pack()
label = Label(frame, text="Name:")
label.grid(row=0, column=0)
entry = Entry(frame)
entry.grid(row=1, column=1)
window.mainloop()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A FEW MORE WIDGETS 312
Parameter Description
row The number of the row to insert the widget into. Row
numbers begin at 0.
column The number of the column to insert the widget into. Col-
umn numbers begin at 0.
rowspan The number of rows the widget occupies. The default is 1.
columnspan The number of columns the widget occupies. The default
is 1.
Figure 14.12: grid() parameters
14.5 A Few More Widgets
To end this chapter, we will look at a few more commonly used widgets.
Text
The Entry widget that we have been using since the start of this chapter
allows for only a single line of text. If we want multiple lines of text, we
use the Text widget instead, as shown here:
Download gui/text.py
from Tkinter import *
def cross(text):
text.insert(INSERT, 'X')
window = Tk()
frame = Frame(window)
frame.pack()
text = Text(frame, height=3, width=10)
text.pack()
button = Button(frame, text="Add", command=lambda: cross(text))
button.pack()
window.mainloop()
Text provides a much richer set of methods than the other widgets we
have seen so far. We can embed images in the text area, put in tags,
select particular lines, and so on. The exercises will give you a chance
to explore its capabilities.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A FEW MORE WIDGETS 313
Figure 14.13: A window with grid layout
Figure 14.14: A window with a text field
Figure 14.15: A checkbutton-based color picker
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A FEW MORE WIDGETS 314
Checkbutton
Checkbuttons, often called checkboxes, have two states, on and off.
When a user clicks a checkbutton, the state changes (see Figure 14.15,
on the previous page). We use a Tkinter mutable variable to keep track
of the user’s selection. Typically, a VarInt variable is used, and the values
1 and 0 indicate on and off, respectively. In the following code, we use
three checkbuttons to create a simpler color picker, and we use the
config method to change the configuration of a widget after it has been
created:
Download gui/checkbutton.py
from Tkinter import *
window = Tk()
frame = Frame(window)
frame.pack()
red = IntVar()
green = IntVar()
blue = IntVar()
for (name, var) in (('R', red), ('G', green), ('B', blue)):
check = Checkbutton(frame, text=name, variable=var)
check.pack(side='left')
def recolor(widget, r, g, b):
color = '#'
for var in (r, g, b):
color += 'FF' if var.get() else '00'
widget.config(bg=color)
label = Label(frame, text='[ ]')
button = Button(frame, text='update',
command=lambda: recolor(label, red, green, blue))
button.pack(side='left')
label.pack(side='left')
window.mainloop()
Menu
The last widget we will look at is Menu.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
A FEW MORE WIDGETS 315
The following code uses this to create the simple text editor shown in
Figure 14.16, on the following page:
Download gui/menu.py
from Tkinter import *import tkFileDialog as dialog
def save(root, text):
data = text.get('0.0', END)
filename = dialog.asksaveasfilename(
parent=root,
filetypes=[('Text', '*.txt')],
title='Save as...')
writer = open(filename, 'w')
writer.write(data)
writer.close()
def quit(root):
root.destroy()
window = Tk()
text = Text(window)
text.pack()
menubar = Menu(window)
filemenu = Menu(menubar)
filemenu.add_command(label='Save', command=lambda : save(window, text))
filemenu.add_command(label='Quit', command=lambda : quit(window))
menubar.add_cascade(label = 'File', menu=filemenu)
window.config(menu=menubar)
window.mainloop()
The program begins by defining two functions: save, which saves the
contents of a text widget, and quit, which closes the application. The
save function uses tkFileDialog to create a standard “Save as...” dialog
box, which will prompt the user for the name of a text file.
After creating and packing the Text widget, the program creates a menu-
bar, which is the horizontal bar into which we can put one or more
menus. It then creates a File menu and adds two menu items to it
called Save and Quit. We then add the File menu to the menu bar and
run mainloop.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
OBJECT -ORIENTED GUIS 316
Figure 14.16: A window with a menu
14.6 Object-Oriented GUIs
The GUIs we have built so far have not been particularly well struc-
tured. Most of the code to construct them has not been modularized
in functions, and they have relied on global variables. We can get away
with this for very small examples, but if we try to build larger applica-
tions this way, they will be difficult to understand and debug.
For this reason, almost all real GUIs are built using classes and objects
that tie models, views, and controllers together in one tidy package.
In the counter shown next, for example, the application’s model is a
member variable of the class Counter called self.state, and its controllers
are the methods upClick and quitClick.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 317
Download gui/oogui.py
from Tkinter import *
class Counter:
'''A simple counter GUI using object-oriented programming.'''
def __init__(self, parent):
'''Create the GUI.'''
# Framework.
self.parent = parent
self.frame = Frame(parent)
self.frame.pack()
# Model.
self.state = IntVar()
self.state.set(1)
# Label displaying current state.
self.label = Label(self.frame, textvariable=self.state)
self.label.pack()
# Buttons to control application.
self.up = Button(self.frame, text='up', command=self.upClick)
self.up.pack(side='left')
self.right = Button(self.frame, text='quit', command=self.quitClick)
self.right.pack(side='left')
def upClick(self):
'''Handle click on 'up' button.'''
self.state.set(self.state.get() + 1)
def quitClick(self):
'''Handle click on 'quit' button.'''
self.parent.destroy()
if __name__ == '__main__':
window = Tk()
myapp = Counter(window)
window.mainloop()
14.7 Summary
In this chapter, we learned the following:
• Most modern programs provide a graphical user interface (GUI)
for displaying information and interacting with users. GUIs are
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 318
built out of widgets, such as buttons, sliders, and text panels; all
modern programming languages provide at least one GUI toolkit.
• Unlike command-line programs, GUI applications are usually
event-driven. In other words, they react to events such as key-
strokes and mouse clicks when and as they occur.
• Experience shows that GUIs should be built using the Model-View-
Controller pattern. The model is the data being manipulated; the
view displays the current state of the data and gathers input from
the user, while the controller decides what to do next.
• Lambda expressions create functions that have no names. These
are often used to define the actions that widgets should take when
users provide input, without requiring global variables.
• Designing usable GUIs is as challenging a craft as designing soft-
ware. Being good at the latter doesn’t guarantee that you can do
the former, but dozens of good books can help you get started.
14.8 Exercises
Here are some exercises for you to try on your own:
1. Write a GUI application with a button labeled “Good-bye.” When
the button is clicked, the window closes.
2. Write a GUI application with a single button. Initially, the button
is labeled 0, but each time it is clicked, the value on the button
increases by 1.
3. What is a more readable way to write the following?
x = lambda(): 3
4. A DNA sequence is a string made up of As, Ts, Cs, and Gs. Write
a GUI application in which a DNA sequence is entered, and when
the Count button is clicked, the number of As, Ts, Cs and Gs are
counted and displayed in the window (see Figure 14.17, on the
next page).
5. In Section 2.6, Local Variables, on page 32, we wrote a function
to convert degrees Fahrenheit to Celsius. Write a GUI application
that looks like Figure 14.18, on the next page.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 319
Figure 14.17: Occurrence of letters in a DNA sequence
Figure 14.18: A temperature conversion GUI
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 320
Figure 14.19: Temperature conversion results
When a value is entered in the text field and the Convert button is
clicked, the value should be converted from Fahrenheit to Celsius
and displayed in the window, as shown in Figure 14.19.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Chapter 15
DatabasesIn earlier chapters, we used files to store data. This is fine for small
problems, but as our data sets become larger and more complex, we
need something that will let us search for data in many different ways,
control who can view and modify the data, and ensure that the data is
correctly formatted. In short, we need a database.
Many different kinds of databases exist. Some are like a dictionary that
automatically saves itself on disk, while others store backup copies of
the objects in a program. The most popular by far, however, are rela-
tional databases, which are at the heart of most large commercial and
scientific software systems. In this chapter, we will introduce the key
concepts behind relational databases and then show you how to per-
form a few common operations. If you would like to know more, there
are thousands of books to turn to; [Feh03] is a good place for newcom-
ers to start.
15.1 The Big Picture
A relational database is a collection of tables, each of which has a fixed
number of columns and a variable number of rows. Each column in a
table has a name and contains values of the same data type, such as
integer or string. Each row, or record, contains values that are related
to each other, such as a particular patient’s name, age, and blood type.
Superficially, each table looks like a spreadsheet or a file with one
record per line (see Section 8.1, One Record per Line, on page 160),
but behind the scenes, the database does a lot of work to keep track of
which values are where and how the tables relate to one another.
Prepared exclusively for Trieu Nguyen
THE BIG PICTURE 322
There are many different brands of database to choose from, including
commercial systems like Oracle, IBM’s DB2, and Microsoft Access, and
open source databases like MySQL and PostgreSQL. Our examples use
one called SQLite. It isn’t fast enough to handle the heavy loads that
sites like Amazon.com experience, but it is free, it is simple to use, and
as of Python 2.5 the standard library includes a module called sqlite3
for working with it.
A database is usually stored in a file, or in a collection of files. These
files are not formatted as plain text—if you open them in an editor, they
will look like garbage, and any changes you make will probably corrupt
the data and make the database unusable. Instead, you must interact
with the database in one of two ways:
• By typing commands into a GUI, just as you type commands into a
Python interpreter. This is good for simple tasks but not for writing
applications of your own.
• By writing programs in Python (or some other language). These
programs import a library that knows how to work with the kind of
database you are using and use that library to create tables, insert
records, and fetch the data you want. Your code can then format
the results in a web page, calculate statistics, or do whatever else
you want.
As shown in Figure 15.1, on the following page, database libraries can
work in two different ways. The simplest is to manipulate the database
directly, just as a program would manipulate a file. In most cases,
though, the library never actually touches the database itself. Instead,
it communicates with a separate program called a database manage-
ment system, or DBMS, which may run on a separate machine and
which is designed to manage connections from dozens or hundreds
of programs at once. Setting up a DBMS isn’t hard, but it isn’t triv-
ial either; the fact that SQLite takes the “direct” approach is another
reason we chose to use it in this chapter.
Python and other languages do their best to hide the disparity between
these two approaches, and between different DBMSs, by having every
database-specific library implement the same application programming
interface, or API. A library’s API is just the set of functions it provides for
programs to call. If two libraries provide the same API, then programs
can switch from one to the other without any code having to be rewrit-
ten. We have already seen this idea in the picture module introduced
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FIRST STEPS 323
SQLCommands
databasePython Program
DBMS(pysqlite)
Figure 15.1: Database architecture
in Section 4.3, Objects and Methods, on page 60, which treated GIF,
JPEG, and PNG images the same way. In the examples in this chapter,
our programs all start with this line:
Download db/db_import.cmd
>>> import sqlite3 as dbapi
which means, “Find version 3 of the SQLite library, and import ver-
sion 2 of the database API from it using the name dbapi.” If we change
our minds later and want to use MySQL as a database, we would just
change this line to this:
Download db/db_mysql.cmd
>>> import MySQLdb as dbapi
and leave the rest of our code alone.
Python’s database API hides some of the differences between different
database systems, but not all of them. To put data into a database
or get information out, we must write commands in a special-purpose
language called SQL, which stands for Structured Query Language and
is pronounced either as “sequel” or as the three letters “S-Q-L.” There
are international standards for SQL, but unfortunately, every database
interprets those standards differently or adds a few “improvements”
that no other database provides. As a result, SQL that works for one
database system may work differently, or not at all, for another. Tools
like SQLAlchemy (http://www.sqlalchemy.org/ and [Cop08]) have been
built to hide the differences between SQL dialects, but they are beyond
the scope of this course.
15.2 First Steps
That’s enough theory; let’s create a database and start doing things
with it. As a running example, we will use the predictions for regional
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FIRST STEPS 324
Region Population
Central Africa 330993
Southeastern Africa 743112
Northern Africa 1037463
Southern Asia 2051941
Asia Pacific 785468
Middle East 687630
Eastern Asia 1362955
South America 593121
Eastern Europe 223427
North America 661157
Western Europe 387933
Japan 100562
Figure 15.2: Estimated world population in 2300
populations in the year 2300 shown in Figure 15.2, which is taken
from http://www.worldmapper.org. (These values are shown graphically
in Figure 15.3, on the next page.)
As promised earlier, we start by telling Python that we want to use sqlite3
by importing the database API:
Download db/db_import.cmd
>>> import sqlite3 as dbapi
Next we must make a connection to our database by calling the data-
base module’s connect method. This method takes one string as a
parameter, which identifies the database we want to connect to. Since
SQLite stores each entire database in a single file on disk, this is just
the path to the file. If the database does not exist, it will be created.
Download db/db_connect.cmd
>>> con = dbapi.connect('population.db')
Once we have a connection, we need to get a cursor. Like the cursor in
your editor, this keeps track of where we are in the database so that
if several programs are accessing the database at the same time, the
database can keep track of who is trying to do what:
Download db/db_cursor.cmd
>>> cur = con.cursor()
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FIRST STEPS 325
Figure 15.3: World populations in 2300 shown graphically
We can now actually start working with the database. The first step is
to create a table to store the population data. To do this, we have to
describe the operation we want in SQL, put that SQL in a string, and
tell the database to execute that string. The general form of the SQL
statement for table creation is as follows:
CREATE TABLE TableName(ColumnName Type, ColumnName Type, ...)
where the names of tables and columns are like the names of variables
in a program and the types are chosen from the types the database
supports (which we will talk about in a couple of paragraph). To cre-
ate a two-column table to store region names as strings and projected
populations as integers, we use this:
Download db/db_create.cmd
>>> cur.execute('CREATE TABLE PopByRegion(Region TEXT, Population INTEGER)')
<sqlite3.dbapi2.Cursor object at 0x00AEEC50>
Our table is called PopByRegion; as you can see, executing the command
returns a cursor object, which in this case we don’t actually need.
The most commonly used data types in SQLite databases are listed in
Figure 15.4, on the following page, along with the corresponding Python
data types. A few of these deserve some more explanation:
• Python stores integers using a single 32-bit word of memory if the
value will fit into it or a more complex multiword structure if it
will not. When fetching values from a database, the sqlite3 library
decides which to use based on the size of those values.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
FIRST STEPS 326
Type Python Equivalent Use
NULL NoneType Means “know nothing about it”
INTEGER int or long Integers
REAL float 8-byte floating-point numbers
TEXT unicode or str Strings of characters
BLOB buffer Binary data
Figure 15.4: SQLite data types
• As we said way back in Section 3.1, Strings, on page 39, Python
normally stores strings using ASCII, which represents each char-
acter using a single byte and includes only those characters com-
mon in English. Python and other (programming) languages use
another scheme called Unicode to represent characters from other
alphabets like Cyrillic, Arabic, Devanagari, and Thai. By default,
sqlite3 represents strings taken from databases as Unicode; we will
see in a moment how to get it to use the more familiar str.
• The term BLOB stands for Binary Large OBject, which to a data-
base means a picture, an MP3, or any other lump of bytes that
isn’t of a more specific type. The Python equivalent is a type we
have not seen before called buffer, which also stores a sequence of
bytes that have no particular predefined meaning. We will not use
BLOBs in our examples, but the exercises will give you a chance
to experiment with them.
After we create a table, our next task is to insert data into it. We do this
one record at a time using the INSERT command, whose general form is
as follows:
INSERT INTO TableName VALUES(Value1, Value2, ...)
As with the parameters to a function call, the values are matched left
to right against the columns. For example, we insert data into the Pop-
ByRegion table like this:
Download db/db_insert.cmd
>>> cur.execute('INSERT INTO PopByRegion VALUES("Central Africa", 330993)')
>>> cur.execute('INSERT INTO PopByRegion VALUES("Southeastern Africa", 743112)')
...
>>> cur.execute('INSERT INTO PopByRegion VALUES("Japan", 100562)')
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
RETRIEVING DATA 327
Notice that the number and type of values in the INSERT statements
matches the number and type of columns in the database table. If we
try to insert a value of a different type than the one declared for the
column, the library will try to convert it, just as it converts the integer
5 to a floating-point number when we do 1.2 + 5. For example, if we insert
the integer 32 into a TEXT column, it will automatically be converted to
“32”; similarly, if we insert a string into an INTEGER column, it is parsed
to see whether it represents a number. If so, the number is inserted.
If the number of values being inserted does not match the number of
columns in the table, the database reports an error, and the data is not
inserted. Surprisingly, though, if we try to insert a value that cannot be
converted to the correct type, such as the string “string” into an INTEGER
field, SQLite will actually do it (though other databases will not).
Saving Changes
After we’ve inserted data into the database or made any other changes,
we must commit those changes using the connection’s commit method:
Download db/db_commit.cmd
>>> con.commit()
Committing to a database is like saving the changes made to a file in a
text editor. Until we do it, our changes are not actually stored and are
not visible to anyone else who is using the database at the same time.
Requiring programs to commit is a form of insurance. If a program
crashes partway through a long sequence of database operations and
commit is never called, then the database will appear as it did before
any of those operations was executed.
15.3 Retrieving Data
Now that we have data in our database, we can start to run queries
to search for data that meets specified criteria. The general form of a
query is as follows:
SELECT ColumnName, ColumnName, ... FROM Table
where TABLE is the name of the table we want to get data from and the
column names specify which values we want. For example, this query
retrieves all the data in the table PopByRegion:
Download db/db_query_allYear2300.cmd
>>> cur.execute('SELECT Region, Population FROM PopByRegion')
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
RETRIEVING DATA 328
Once the database has executed this query for us, we can access the
results one record at a time by calling the cursor’s fetchone method,
just as we can read one line at a time from a file using readline:
Download db/db_fetchone.cmd
>>> print cur.fetchone()
(u'Central Africa', 330993)
The fetchone method returns each record as a tuple (Section 5.9, Other
Kinds of Sequences, on page 99) whose elements are in the order spec-
ified in the query. If there are no more records, fetchone returns None.
By default, TEXT values in the database are returned as Unicode strings
(as indicated by the u prefix in front of the string ’Central Africa’). We can
tell sqlite3 to return strings as type str instead by assigning the type str
to the cursor’s text_factory member:
Download db/db_fetchone_2.cmd
con.text_factory = str
>>> print cur.fetchone()
('Northern Africa', 1037163)
Just as files have a readlines method to get all the lines in a file at
once, database cursors have a fetchall method that returns all the data
produced by a query as a list of tuples:
Download db/db_query_allYear2300_2.cmd
>>> print cur.fetchall()
[('Souteastern Africa', 743112), ('Asia Pacific', 785468), ('Middle East',
687630), ('Eastern Asia', 1362955), ('South America', 593121), ('Eastern
Europe', 223427), ('North America', 661157), ('Western Europe', 387933),
('Japan', 100562)]
Notice that the tuples are not sorted in any way. Like a dictionary or
a set (Chapter 9, Sets and Dictionaries, on page 185), a database store
records in whatever order it think is most efficient. To put the data in a
particular order, we could sort the list returned by fetchall. However, it
is more efficient to get the database to do the sorting for us by adding
an ORDER BY clause to the query like this:
Download db/db_query_sort.cmd
>>> cur.execute('SELECT Region, Population FROM PopByRegion ORDER BY Region')
>>> cur.fetchall()
[('Asia Pacific', 785468), ('Central Africa', 330993), ('Eastern
Asia', 1362955), ('Eastern Europe', 223427), ('Japan', 100562),
('Middle East', 687630), ('North America', 661157), ('Northern
Africa', 1037463), ('South America', 593121), ('Southeastern Africa',
743112), ('Southern Asia', 2051941), ('Western Europe', 387933)]
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
RETRIEVING DATA 329
By changing the column name after the phrase ORDER BY, we can change
the way the database sorts. As the following code shows, we can also
specify whether we want values sorted in ascending (ASC) or descending
(DESC) order:
Download db/db_query_sort_2.cmd
>>> cur.execute('SELECT Region, Population FROM PopByRegion
ORDER BY Population DESC')
>>> cur.fetchall()
[('Southern Asia', 2051941), ('Eastern Asia', 1362955), ('Northern
Africa', 1037463), ('Asia Pacific', 785468), ('Southeastern Africa',
743112), ('Middle East', 687630), ('North America', 661157), ('South
America', 593121), ('Western Europe', 387933), ('Central Africa',
330993), ('Eastern Europe', 223427), ('Japan', 100562)]
Rather than getting all columns, we can specify one or more columns
by name. We can also use * to indicate that we want all columns, just
as we would use import * to import all of the contents of a module:
Download db/db_more_queries.cmd
>>> cur.execute('SELECT Region FROM PopByRegion')
[('Central Africa',), ('Southeastern Africa',), ('Northern Africa',),
('Southern Asia',), ('Asia Pacific',), ('Middle East',), ('Eastern
Asia',), ('South America',), ('Eastern Europe',), ('North America',),
('Western Europe',), ('Japan',)]
>>> cur.execute('SELECT * FROM PopByRegion')
[('Central Africa', 330993), ('Southeastern Africa', 743112),
('Northern Africa', 1037463), ('Southern Asia', 2051941), ('Asia
Pacific', 785468), ('Middle East', 687630), ('Eastern Asia', 1362955),
('South America', 593121), ('Eastern Europe', 223427), ('North
America', 661157), ('Western Europe', 387933), ('Japan', 100562)]
Query Conditions
Much of the time, we want only some of the data in the database. (Think
about what would happen if you asked Google for all of the web pages
it had stored.) We can select a subset of the data by using the keyword
WHERE to specify conditions that the rows we want must satisfy. For
example, we can get the regions with populations greater than 1 million
using the greater-than operator:
Download db/db_where.cmd
>>> cur.execute('SELECT Region FROM PopByRegion WHERE Population > 1000000')
>>> print cur.fetchall()
[('Northern Africa',), ('Southern Asia',), ('Eastern Asia',))]
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
UPDATING AND DELETING 330
Operator Description
= Equal to
!= Not equal to
> Greater than
< Less than
>= Greater than or equal to
<= Less than or equal to
Figure 15.5: SQL relational operators
The relational operators that may be used with WHERE are listed in Fig-
ure 15.5. Not surprisingly, they are the same as the ones that Python
and other programming languages provide.
As well as these relational operators, we can also use the Boolean oper-
ators AND, OR, and NOT. To get a list of regions with populations greater
than 1 million that have names that come before the letter L in the
alphabet, we would use this:
Download db/db_where_2.cmd
>>> cur.execute('SELECT Region FROM PopByRegion
WHERE Population > 1000000 AND Region < "L"')
>>> print cur.fetchall()
[('Eastern Asia',))]
WHERE conditions are always applied row by row—they cannot be used
to compare two or more rows. We will see how to do that in Section 15.7,
Using Joins to Combine Tables, on page 334.
15.4 Updating and Deleting
Data often changes over time, so we need to be able to change the infor-
mation stored in databases. To do that, we use the UPDATE command,
as shown here:
Download db/db_update.cmd
>>> cur.execute('SELECT * FROM PopByRegion WHERE Region = "Japan"')
>>> cur.fetchone()
('Japan', 100562)
>>> cur.execute('UPDATE PopByRegion SET Population = 100600
WHERE Region = "Japan"')
>>> cur.execute('SELECT * FROM PopByRegion WHERE Region = "Japan"')
>>> cur.fetchone()
('Japan', 100600)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TRANSACTIONS 331
We can also delete records from the database:
Download db/db_delete.cmd
>>> cur.execute('DELETE FROM PopByRegion WHERE Region < "L"')
>>> cur.execute('SELECT * FROM PopByRegion');
>>> cur.fetchall()
[('Southeastern Africa', 743112), ('Northern Africa', 1037463),
('Southern Asia', 2051941), ('Middle East', 687630), ('South America',
593121), ('North America', 661157), ('Western Europe', 387933)]
In both cases, all records that meet the WHERE condition are affected. If
we do not include a WHERE condition, then all rows in the database are
updated or removed. Of course, we can always put records back into
the database:
Download db/db_delete_2.cmd
>>> cur.execute('INSERT INTO PopByRegion VALUES ("Japan", 100562)')
To remove an entire table from the database, we can use the DROP
command:
DROP TABLE TableName
For example, if we no longer want the table PopByRegion, we would exe-
cute this:
Download db/db_drop.cmd
>>> cur.execute('DROP TABLE PopByRegion');
When a table is dropped, all the data it contained is lost. You should be
very, very sure you want to do this (and even then, it’s probably a good
idea to make a backup copy of the database before deleting any sizable
tables).
15.5 Transactions
Database operations are almost always grouped into a transaction. No
operation in a transaction can be committed unless every single one
can be successfully committed in sequence. If an operations fails, the
transaction must be rolled back. That causes all the operations in the
transaction to be undone. Using transactions ensures the database
doesn’t contain half-baked results. Databases create transactions auto-
matically. As soon as you try to start an operation, it becomes part
of a transaction. When you commit the transaction successfully, the
changes becomes permanent, and the database creates a new one.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
TRANSACTIONS 332
Imagine a library that may have multiple copies of the same book. It
uses a computerized system to track its books by their ISBN number.
Whenever a patron signs out a book, the following code is executed by
one of the library computers:
Download db/db_transaction_1.cmd
cur.execute('SELECT SignedOut FROM Books WHERE ISBN = "%s"' % isbn)
signedOut = cur.fetchone()[0]
cur.execute('UPDATE Books SET SignedOut = %d
WHERE ISBN = "%s"' % (signedOut + 1, isbn))
cur.commit()
When a patron returns a book, the reverse happens:
Download db/db_transaction_2.cmd
cur.execute('SELECT SignedOut FROM Books WHERE ISBN = "%s"' % isbn)
signedOut = cur.fetchone()[0]
cur.execute('UPDATE Books SET SignedOut = %d
WHERE ISBN = "%s"' % (signedOut - 1, isbn))
cur.commit()
What if the library had two computers that handled book signouts and
returns? Both computers connect to the same database. What would
happen if one patron tried to return a copy of Gray’s Anatomy, while
another was signing out a different copy of the same book at the exact
same time? Here’s one possibility:
Download db/db_transaction_3.cmd
Computer A: cur.execute('SELECT SignedOut FROM Books WHERE ISBN = "%s"' % isbn)
Computer A: signedOut = cur.fetchone()[0]
Computer B: cur.execute('SELECT SignedOut FROM Books WHERE ISBN = "%s"' % isbn)
Computer B: signedOut = cur.fetchone()[0]
Computer A: cur.execute('UPDATE Books SET SignedOut = %d
WHERE ISBN = "%s"' % (signedOut + 1, isbn))
Computer A: cur.commit()
Computer B: cur.execute('UPDATE Books SET SignedOut = %d
WHERE ISBN = "%s"' % (signedOut - 1, isbn))
Computer B: cur.commit()
Notice that Computer B counts the number of signed-out copies before
Computer A updates the database. After Computer A commits its
changes, the value that Computer B fetched is no longer accurate. If
Computer B were allowed to commit its changes, the library database
would account for more books than the library actually has!
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USING NULL FOR MISSING DATA 333
Fortunately, databases can detect such a situation and would prevent
Computer B from committing its transaction.
15.6 Using NULL for Missing Data
In the real world, we often don’t have all the data we want. We might be
missing the time at which an experiment was performed or the postal
code of a patient being given a new kind of treatment. Rather than leave
what we do know out of the database, we may choose to insert it and
use the value NULL to represent the missing values. For example, if there
is a region whose population we don’t know, we could insert this into
our database:
Download db/db_null.cmd
>>> cur.execute('INSERT INTO PopByRegion VALUES ("Mars", NULL)')
On the other hand, we probably don’t ever want a record in the data-
base that has a NULL region name. We can prevent this from ever hap-
pening, stating that the column is NOT NULL when the table is created:
Download db/db_null_2.cmd
>>> cur.execute('CREATE TABLE Test (Region TEXT NOT NULL, Population INTEGER)')
Now when we try to insert a NULL region into our new Test table, we get
an error message:
Download db/db_null_3.cmd
>>> cur.execute('INSERT INTO Test VALUES (NULL, 456789)')
Traceback (most recent call last):
File "<string>", line 1, in <string>
sqlite3.dbapi2.IntegrityError: Test.Region may not be NULL
Stating that the value must not be NULL is not always necessary, and
imposing such a constraint may not be reasonable in some cases.
Rather than using NULL, it may sometimes be more appropriate to use
the value zero, an empty string, or false. You should do so in cases
where you know something about the data and use NULL only in cases
where you know nothing at all about it.
In fact, some experts recommend not using NULL at all because its
behavior is counterintuitive (at least until you’ve retrained your intu-
ition). The general rule is that operations involving NULL produce NULL
as a result; the reasoning is that if the computer doesn’t know what
one of the operation’s inputs is, it can’t know what the output is either.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USING JOINS TO COMBINE TABLES 334
Adding a number to NULL therefore produces NULL, no matter what the
number was, and multiplying by NULL also produces NULL.
Things are more complicated with logical operations. The expression
NULL OR 1 produces 1, rather than NULL, because of the following:
• If the first argument was false (or 0, or the empty string, or some
equivalent value), the result would be 1.
• If the first argument was true (or nonzero, or a nonempty string)
the result would also be 1.
The technical term for this is three-valued logic. In SQL’s view of the
world, things aren’t just true or false—they can be true, false, or un-
known, and NULL represents the latter. Unfortunately, different data-
bases interpret ambiguities in the SQL standard in different ways, so
their handling of NULL is not consistent. NULL should therefore be used
with caution and only when other approaches won’t work.
15.7 Using Joins to Combine Tables
When designing a database, it often makes sense to divide data between
two or more tables. For example, if we are maintaining a database of
patient records, we would probably want at least four tables: one for the
patient’s personal information (such as their name date of birth), a sec-
ond to keep track of their appointments, a third for information about
the doctors who are treating them, and a fourth for information about
the hospitals those doctors work in (see Figure 15.6, on the following
page). We could store all of this in one table, as shown Figure 15.7,
on the next page, but then a lot of information would be needlessly
duplicated.
If we divide information between tables, though, we need some way to
pull that information back together. For example, if we want to know
the hospitals at which a patient has had appointments, we need to
combine data from all four tables to find out:
• What appointments the patient has had
• Which doctor each appointment was with
• Which hospital that doctor works at
The right way to do this in a relational database is to use a join. As the
name suggests, a join combines information from two or more tables to
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USING JOINS TO COMBINE TABLES 335
PatientAppointment
DoctorHospital
patient
doctor
date
name
birthday
name
address
name
hospital
Figure 15.6: Dividing data between tables
Patient-Doctor-Appointment-Hospital
patient doctor datebirthday hospital address
Alice
Alice
Alice
Zack
Zack
1978/04/02
1964/12/15
1978/04/02
1978/04/02
1964/12/15
Rajani
Newton
Nianiaris
Newton
Vaz
2008/09/01 Central
East
Central
East
East
2008/09/14
2008/10/04
2008/09/18
2008/11/01
52 Walnut St.
8 Elm St.
52 Walnut St.
8 Elm St.
8 Elm St.
Figure 15.7: A bad database design
create a new set of records, each of which can contain some or all of
the information in the tables involved.
To begin, let’s add another table that contains the names of countries,
the regions that they are in, and their populations:
Download db/db_add_tables.cmd
>>> cur.execute('CREATE TABLE PopByCountry(Region TEXT, Country TEXT,
Population INTEGER)')
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USING JOINS TO COMBINE TABLES 336
and then insert data into the new table:
Download db/db_insert_2.cmd
>>> cur.execute('INSERT INTO PopByCountry VALUES("Eastern Asia", "China",
1285238)')
Inserting data one row at a time like this requires a lot of typing. It is
simpler to make a list of tuples to be inserted and write a loop that
inserts the values from these tuples one by one:
Download db/db_insert_3.cmd
>>> countries = [("Eastern Asia", "DPR Korea", 24056), ("Eastern Asia",
"Hong Kong (China)", 8764), ("Eastern Asia", "Mongolia", 3407), ("Eastern
Asia", "Republic of Korea", 41491), ("Eastern Asia", "Taiwan", 1433),
("North America", "Bahamas", 368), ("North America", "Canada", 40876),
("North America", "Greenland", 43), ("North America", "Mexico", 126875),
("North America", "United States", 493038)]
>>> for c in countries:
... cur.execute('INSERT INTO PopByCountry VALUES (?, ?, ?)', (c[0], c[1], c[2]))
...
>>> con.commit()
This time, the call to execute has two arguments. The first is the SQL
command with question marks as placeholders for the values we want
to insert. The second is a tuple of values, which the database matches
up against the question marks from left to right when it executes the
command.
Now that we have two tables in our database, we can use joins to com-
bine the information they contain. There are several types of joins; we
will begin with inner joins, which involve the following:
1. Constructing the cross product of the tables
2. Discarding rows that do not meet the selection criteria
3. Selecting columns from the remaining rows
These steps are shown graphically in Figure 15.8, on the following page.
First, all combinations of all rows in the tables are combined, which
makes the cross product. Second, the selection criteria specified by
WHERE is applied, and rows that don’t match are removed. Finally, the
selected columns are kept, and all others are discarded.
In an earlier query, we retrieved the names of regions with projected
populations greater than 1 million. Using an inner join, we can get the
names of the countries that are in those regions.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USING JOINS TO COMBINE TABLES 337
Eastern Asia
North America
1362955
661157
Eastern Asia
North America
1362955
661157
Eastern Asia
North America
Mongolia
Greenland
3407
43
North America Greenland 43
Eastern Asia Mongolia 3407
Eastern Asia
North America
1362955
661157
Eastern Asia
North America
Mongolia
Greenland
3407
43
Eastern Asia 1362955 Eastern Asia Mongolia 3407
Eastern Asia
North America
1362955
661157
Eastern Asia
North America
Mongolia
Greenland
3407
43
PopByRegion PopByCountry
Keep rows where PopByRegion.Region = PopByCountry.Region2a
Keep rows where PopByRegion.Population > 10000002b
Compute cross product1
Keep columns PopByRegion.Region and PopByCountry.Country3
Figure 15.8: Inner joins in action
The query and its result look like this:
Download db/db_inner_join.cmd
>>> cur.execute('''
SELECT PopByRegion.Region, PopByCountry.Country
FROM PopByRegion INNER JOIN PopByCountry
WHERE (PopByRegion.Region = PopByCountry.Region)
AND (PopByRegion.Population > 1000000)
''')
>>> print cur.fetchall()
[('Eastern Asia', 'China'), ('Eastern Asia', 'DPR Korea'),
('Eastern Asia', 'Hong Kong (China)'), ('Eastern Asia', 'Mongolia'),
('Eastern Asia', 'Republic of Korea'), ('Eastern Asia', 'Taiwan')]
To understand what this query is doing, we can analyze it in terms of
the three steps outlined earlier:
1. Combine every row of PopByRegion with every row of PopByCountry.
PopByRegion has two columns and twelve rows, while PopByCountry
has three columns and eleven rows, so this produces a temporary
table with five columns and 132 rows (see Figure 15.9, on the next
page).
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
USING JOINS TO COMBINE TABLES 338
Central Africa
Southeastern Africa
330993
743112
Northern Africa 1037463
Southern Asia
Asia Pacific
2051941
785468
Middle East 687630
Eastern Asia
South America
1362955
593121
Eastern Europe 223427
DPR Korea
Hong Kong (China)
North America
Western Europe
Japan
661157
387933
100562
Eastern Asia
Eastern Asia
24056
8764
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
DPR KoreaEastern Asia 24056
Central Africa
Southeastern Africa
330993
743112
Northern Africa 1037463
Southern Asia
Asia Pacific
2051941
785468
Middle East 687630
Eastern Asia
South America
1362955
593121
Eastern Europe 223427
North America
Western Europe
Japan
661157
387933
100562
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
Hong Kong (China)Eastern Asia 8764
... ... ...... ...
Figure 15.9: An inner join in progress
2. Discard rows that do not meet the selection criteria. The join’s
WHERE clause specifies two of these: the region taken from Pop-
ByRegion must be the same as the region taken from PopByCountry,
and the region’s population must be greater than 1 million. The
first criterion ensures that we don’t look at records that combine
countries in North America with regional populations in East Asia;
the second filters out information about countries in region whose
populations are less than our threshold.
3. Finally, we select the region and country names from the rows that
have survived.
To find the regions where one country accounts for more than 10 per-
cent of the region’s overall population, we would also need to join the
two tables.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
KEYS AND CONSTRAINTS 339
Download db/db_duplicates.cmd
>>> cur.execute('''
SELECT PopByRegion.Region
FROM PopByRegion INNER JOIN PopByCountry
WHERE (PopByRegion.Region = PopByCountry.Region)
AND ((PopByCountry.Population * 1.0) / PopByRegion.Population > 0.10)''')
>>> print cur.fetchall()
[('Eastern Asia',), ('North America',), ('North America',)]
We use multiplication and division in our WHERE condition to calcu-
late the percentage of the region’s population by country as a floating-
point number. The resulting list contains duplicates, since more than
one Eastern Asian country accounts for more than 10 percent of the
region’s population. To remove the duplicates, we add the keyword DIS-
TINCT to the query:
Download db/db_distinct.cmd
>>> cur.execute('''
SELECT PopByRegion.Region
FROM PopByRegion INNER JOIN PopByCountry
WHERE (PopByRegion.Region = PopByCountry.Region)
AND ((PopByCountry.Population * 1.0) / PopByRegion.Population > 0.10)''')
>>> print cur.fetchall()
[('Eastern Asia',), ('North America',)]
15.8 Keys and Constraints
Our query in the previous section relied on the fact that our regions
and countries were uniquely identified by their names. A column in a
table that is used this way is called a key. Ideally, a key’s values should
be unique, just like the keys in a dictionary. We can tell the database
to enforce this constraint by adding a PRIMARY KEY clause when we cre-
ate the table. For example, when we created the PopByRegion table, we
should have specified the primary key:
Download db/db_primary_key.cmd
>>> cur.execute('CREATE TABLE PopByRegion (Region TEXT NOT NULL,
Population INTEGER NOT NULL, PRIMARY KEY (Region))');
Just as a key in a dictionary can be made up of multiple values, the
primary key for a database table can consist of multiple columns.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
KEYS AND CONSTRAINTS 340
The following code uses the CONSTRAINT keyword to specify that no two
entries in the table being created will ever have the same values for
region and country:
Download db/db_constraint.cmd
>>> cur.execute('''
CREATE TABLE PopByCountry(
Region TEXT NOT NULL,
Country TEXT NOT NULL,
Population INTEGER NOT NULL,
CONSTRAINT Country_Key PRIMARY KEY (Region, Country))
''')
A table can also contain one or more foreign keys. As the name sug-
gests, these are values that aren’t guaranteed to be unique in that table
but that are (unique) keys in another table. Going back to the example
of patients, appointments, doctors, and hospitals, suppose the hospi-
tal’s name is the primary key in the HOSPITAL table. The DOCTOR table
could then have a column called HOSPITAL_NAME to identify where the
doctor worked. This column would be a foreign key, since its values
wouldn’t be unique in DOCTOR (many doctors could work at the same
hospital) but would be unique in the HOSPITAL table.
In practice, most database designers do not use real names as primary
keys. Instead, they usually create a unique integer ID for each “thing”
in the database, such as a driver’s license number or a patient ID.
This is partly done for efficiency’s sake—integers are faster to sort and
compare than strings—but the real reason is that it is a simple way to
deal with hospitals (or people) that have the same name. There are a
lot of Jane Smiths in the world; using that name as a primary key in a
database is almost guaranteed to lead to confusion. Giving each person
a unique ID, on the other hand, ensures that they can be told apart.
Using unique IDs in this way also makes some operations much easier
to implement. For example, imagine what would happen in a database
Dusing a hospital’s name as a primary key if the hospital changed its
name. Someone would have to write code that found every use of that
hospital’s name and changed it to the new value. If the hospital was
identified by an integer hospital ID, then renaming the hospital would
be as simple as replacing one string with another—the doctor records
that referred to the hospital would not have to be updated.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ADVANCED FEATURES 341
Aggregate Function Description
AVG Average of the values
MIN Minimum value
MAX Maximum value
COUNT Number of non-null values
SUM Sum of the values
Figure 15.10: Aggregate functions
15.9 Advanced Features
The SQL we have seen so far is powerful enough for many everyday
tasks, but other questions require more powerful tools. This chapter
introduces a handful and shows when and how they are useful. If you
need to do things that are even more complicated, you may want to add
[Bea05] and [VH07] to your reading list.
Aggregation
Our next task is to calculate the total projected world population for
the year 2300. We will do this by adding up the values in PopByRegion’s
Population column using the SQL aggregating function SUM:
Download db/db_aggregate.cmd
>>> cur.execute('SELECT SUM (Population) FROM PopByRegion')
>>> cur.fetchone()
(8965762,)
SQL provides many other aggregation functions (see Figure 15.10). All
of these are associative; that is, the result doesn’t depend on the order
of operations.1 This ensures that the result doesn’t depend on the order
in which records are pulled out of tables, which is something that only
the database knows.
Grouping
What if we only had the table PopByCountry and wanted to find the pro-
jected population for each region? We could get the table’s contents into
a Python program using SELECT * and then loop over them to add them
1. Addition and multiplication are associative, since 1 + (2 + 3) produces the same
results as (1 + 2) + 3, and 4 * (5 * 6) produces the same result as (4 * 5) * 6. By contrast,
subtraction is not: 1 - (2 - 3) is not the same thing as (1 - 2) - 3.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ADVANCED FEATURES 342
up by region, but again, it is simpler and more efficient to have the
database do the work for us. In this case, we use SQL’s GROUP BY to
collect results into subsets:
Download db/db_group.cmd
>>> cur.execute('SELECT SUM (Population) FROM PopByCountry GROUP BY Region')
>>> cur.fetchall()
[(1364389,), (661200,)]
Since we have asked the database to construct groups by Region and
there are two distinct values in this column in the table, the database
divides the records into two subsets. It then applies the SUM function to
each group separately to give us the projected populations of Eastern
Asia and North America (see Figure 15.11, on the next page).
Self-Joins
SQL is powerful, but it isn’t always obvious how to translate a simple
question into selects, joins, and Boolean conditions. Once you have
seen a few examples of common cases, though, it is easy to apply the
pattern in other situations.
A common example is the problem of comparing a table’s values to
themselves. Suppose that we want to find pairs of countries whose
populations are close to one another—say, within 1,000 of one another.
Our first attempt might look like this:
Download db/db_self1.cmd
>>> cur.execute('SELECT Country FROM PopByCountry
WHERE (ABS(Population - Population) < 1000)')
>>> cur.fetchall()
[('China',), ('DPR Korea',), ('Hong Kong (China)',), ('Mongolia',),
('Republic of Korea',), ('Taiwan',), ('Bahamas',), ('Canada',),
('Greenland',), ('Mexico',), ('United States',)]
The output is definitely not what we want, and there are two reasons
why. First, the phrase SELECT Country is going to return only one coun-
try per record, but we want pairs of countries. Second, the expression
ABS(Population - Population) is always going to return zero; since we’re
subtracting each country’s population from itself, every results will be
less than 1,000, so the names of all the countries in the table will be
returned.
What we actually want to do is compare the population in one row with
the populations in the other rows. To do this, we need to join PopBy-
Country with itself (see Figure 15.12, on page 344) using an INNER JOIN.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ADVANCED FEATURES 343
PopByCountry
North America
North America
Bahamas
Mexico
368
126875
Eastern Asia
Eastern Asia
Mongolia
Republic of Korea
3407
41491
Eastern Asia
Eastern Asia
Mongolia
Republic of Korea
3407
41491
44898
North America 127243
Eastern Asia 44898
North America Bahamas
Mexico
368
126875
127243
North America
Group by Region and compute SUM of Population1
Keep selected columns2
Figure 15.11: Summation
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ADVANCED FEATURES 344
PopByCountry
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
North America
North America
Canada
United States
40876
493038
Eastern Asia Taiwan 1433
PopByCountry cross joined with itself
Figure 15.12: Joining a table to itself
This will result in the rows for each pair of countries being combined
into a single row with six columns: two regions, two countries, and two
populations. To tell them apart, we have to give the two instances of
the PopByCountry table temporary names (in this case A and B) to each
table:
Download db/db_self_join.cmd
>>> cur.execute('''
SELECT A.Country, B.Country
FROM PopByCountry A INNER JOIN PopByCountry B
WHERE (ABS(A.Population - B.Population) <= 1000)
AND (A.Country != B.Country)''')
>>> cur.fetchall()
[('Republic of Korea', 'Canada'), ('Bahamas', 'Greenland'), ('Canada',
'Republic of Korea'), ('Greenland', 'Bahamas')]
Notice that we have used ABS to get the absolute value of the population
difference. If we simply wrote this:
(A.Population - B.Population) <= 1000
then our results would have included pairs like (’Greenland’, ’China’),
because every negative difference is less than 1,000. If we want each
pair of countries to appear only once, we could rewrite the second half
of the condition as follows:
A.Country < B.Country
which would rule out half of each duplicated pair of countries.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
ADVANCED FEATURES 345
Nested Queries
Up to now, our queries have involved only one SELECT command. Since
the result of every query looks exactly like a table with a fixed number
of columns and some number of rows, we can run a second query on
the result, that is, run a SELECT on the result of another SELECT, rather
than directly on the database’s tables. Such queries are called nested
queries and are analogous to having one function call another, which
might in turn call a third.
To see why we would want to do this, let’s try to write a query on the
PopByCountry table to get the regions that do not have a country with a
population of exactly 8,764. Our first attempt looks like this:
Download db/db_nested_1.cmd
>>> cur.execute('''
SELECT DISTINCT Region
FROM PopByCountry
WHERE (PopByCountry.Population != 8764)
''')
>>> cur.fetchall()
[('Eastern Asia',), ('North America',)]
This result is wrong—Hong Kong has a projected population of 8,764,
so Eastern Asia should not have been returned. Because there are other
countries in Eastern Asia whose populations are not 8,764, though,
Eastern Asia was included in the final results.
Let’s rethink our strategy. What we have to do is to find out which
regions include countries with a population 8,764 and then exclude
those regions from our final result—basically, find the regions that fail
our condition and subtract them from the set of all countries (see Fig-
ure 15.13, on the next page). The first step is to get those regions that
have countries with a population of 8,764:
Download db/db_nested_2.cmd
>>> cur.execute('''
SELECT DISTINCT Region
FROM PopByCountry
WHERE (PopByCountry.Population = 8764)
''')
>>> cur.fetchall()
[('Eastern Asia',)
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
SUMMARY 346
Eastern Asia
North America
Eastern Asia
North America
SELECT DISTINCT Region
FROM PopByCountry
(SELECT DISTINCT Region
FROM PopByCountry
WHERE (PopByCountry.Population = 8764))
WHERE Region NOT IN (-
=
Figure 15.13: Nested negation
Now we want to get the names of regions that were not in the results of
our first query. To do this, we will use a WHERE condition and NOT IN:
Download db/db_nested_3.cmd
>>> cur.execute('''
SELECT DISTINCT Region
FROM PopByCountry
WHERE Region NOT IN
(SELECT DISTINCT Region
FROM PopByCountry
WHERE (PopByCountry.Population = 8764))
''')
>>> cur.fetchall()
[('North America',)]
This time we got what we were looking for. Nested queries are often
used for situations like this one where negation is involved.
15.10 Summary
In this chapter, we learned the following:
• Most large applications store information in relational databases.
A database is made up of tables, each of which stores logically
related information. A table has one or more columns, each of
which has a name and a type, and zero or more rows, or records.
In most tables, each row can be identified by a unique key, which
consists or one or more of the values in the row.
• Commands to put data into databases, or get data out, are writ-
ten in a specialized language called SQL. SQL is supposed to be
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 347
a standard, but there are significant differences in how different
databases implement it.
• SQL commands can be sent to databases interactively from GUIs
or command-line tools, but for larger jobs, it is more common to
write programs that create SQL and process the results. Since
SQL’s data types are not always exactly the same as those in the
language used to write the program, it may be necessary to trans-
late from one to the other.
• Changes made to a database do not actually take effect until they
are committed. This ensures that if two or more programs are
working with a database at the same time, it will always be in a
consistent state. However, it also means that operations in one
program can fail because of something that another program is
doing.
• SQL queries must specify the table(s) and column(s) that values
are to be taken from. They may also specify Boolean conditions
those values must satisfy and the ordering of results.
• Simple queries work on one row at a time, but programs can join
tables to combine values from different rows. Queries can also
group and aggregate rows to calculate sums, averages, and other
values.
• Databases can use the special value NULL to represent missing
information. However, it must be used with caution, since opera-
tions on NULL values do not behave in the same way as operations
on “real” values.
15.11 Exercises
Here are some exercises for you to try on your own:
1. In this exercise, you will create a table to store the population and
land area of the Canadian provinces and territories according to
the 2001 Census. Our data is taken from http://www12.statcan.ca/
english/census01/home/index.cfm.
a) Create a new database called census.db.
b) Make a database table called Density that will hold the name
of the province or territory (TEXT), the population (INTEGER), and
the land area (REAL).
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 348
Province/Territory Population Land Area
Newfoundland and Labrador 512930 370501.69
Prince Edward Island 135294 5684.39
Nova Scotia 908007 52917.43
New Brunswick 729498 71355.67
Quebec 7237479 1357743.08
Ontario 11410046 907655.59
Manitoba 1119583 551937.87
Saskatchewan 978933 586561.35
Alberta 2974807 639987.12
British Columbia 3907738 926492.48
Yukon Territory 28674 474706.97
Northwest Territories 37360 1141108.37
Nunavut 26745 1925460.18
Figure 15.14: 2001 Canadian census data
c) Insert the data from Figure 15.14.
d) Display the contents of the table.
e) Display the populations.
f) Display the provinces that have populations of less than 1
million.
g) Display the provinces that have populations less than 1 mil-
lion or greater than 5 million.
h) Display the provinces that do not have populations less than
1 million or greater than 5 million.
i) Display the populations of provinces that have a land area
greater than 200,000 square kilometers.
j) Display the provinces along with their population densities
(population divided by land area).
2. For this exercise, add a new table called Capitals to the database.
Capitals has three columns—province/territory (TEXT), capital (TEXT),
and population (INTEGER)—and it holds the data shown in Fig-
ure 15.15, on the next page.
a) Retrieve the contents of the table.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 349
Province/Territory Capital Population
Newfoundland and Labrador St. John’s 172918
Prince Edward Island Charlottetown 58358
Nova Scotia Halifax 359183
New Brunswick Fredericton 81346
Quebec Quebec 682757
Ontario Toronto 4682897
Manitoba Winnipeg 671274
Saskatchewan Regina 192800
Alberta Edmonton 937845
British Columbia Victoria 311902
Yukon Territory Whitehorse 21405
Northwest Territories Yellowknife 16541
Nunavut Iqaluit 5236
Figure 15.15: 2001 Canadian census data: capital city populations
b) Retrieve the populations of the provinces and capitals (in a
list of tuples of the form (province population, capital popula-
tion)).
c) Retrieve the land area of the province whose capitals have
populations greater than 100,000.
d) Retrieve the provinces with land densities less than 2 people
per square kilometer and capital city populations more than
500,000.
e) Retrieve the total land area of Canada.
f) Retrieve the average capital city population.
g) Retrieve the lowest capital city population.
h) Retrieve the highest province/territory population.
i) Retrieve the provinces that have land densities within 0.5 of
each other.
3. Write a Python program that creates a new database and executes
the following SQL statements. How do the results of the SELECT
statements differ from what you would expect Python itself to do?
Why?
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
EXERCISES 350
CREATE TABLE Numbers(Val INTEGER)
INSERT INTO Numbers Values(1)
INSERT INTO Numbers Values(2)
SELECT * FROM Numbers WHERE 1/0
SELECT * FROM Numbers WHERE 1/0 AND Val > 0
SELECT * FROM Numbers WHERE Val > 0 AND 1/0
4. A friend of yours has written a run_query function that will execute
a single SQL query against a database and return the results,
doing all of the required setup and teardown automatically:
def run_query(db, query):
'''Return the results of running the given query on database db.'''
con = sqlite.connect(db)
cur = con.cursor()
cur.execute(query)
data = cur.fetchall()
cur.close()
con.close()
return data
# Example of use.
run_query(db, 'SELECT * FROM Precipitation')
Modify this function so that it will use a tuple of arguments if one
is provided so that both the previous call and the following one
will work:
run_query(db, 'SELECT City, Temp, Snow FROM Precipitation
WHERE Snow >= (?) and Temp > (?)', (s,t))
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
Appendix A
Bibliography
[AW06] Mike Andrews and James A. Whittaker. How to Break Web
Software. Addison-Wesley, Boston, MA, 2006.
[Bea05] Alan Beaulieu. Learning SQL. O’Reilly & Associates, Inc,
Sebastopol, CA, 2005.
[Cop08] Rick Copeland. Essential SQLAlchemy. O’Reilly & Asso-
ciates, Inc, Sebastopol, CA, 2008.
[DEM02] Allen Downey, Jeff Elkner, and Chris Meyers. How to Think
Like a Computer Scientist: Learning with Python. Green Tea
Press, Needham, MA, 2002.
[Feh03] Chris Fehily. SQL: Visual QuickStart Guide. Peachpit Press,
Berkeley, CA, 2003.
[GL07] Michael H. Goldwasser and David Letscher. Object-Oriented
Programming in Python. Prentice Hall, Englewood Cliffs, NJ,
2007.
[Guz04] Mark Guzdial. Introduction to Computing and Programming
in Python: A Multimedia Approach. Prentice Hall, Englewood
Cliffs, NJ, 2004.
[Hoc04] Roger R. Hock. Forty Studies That Changed Psychology.
Prentice Hall, Englewood Cliffs, NJ, 2004.
[Hynnd] R. J. Hyndman. Time series data library.
http://www.robjhyndman.com/TSDL, n.d. Accessed on 19
September 2006.
Prepared exclusively for Trieu Nguyen
APPENDIX A. BIBLIOGRAPHY 352
[Joh07] Jeff Johnson. GUI Bloopers 2.0: Common User Interface
Design Don’ts and Dos. Morgan Kaufmann Publishers, San
Francisco, 2007.
[KS04] Marja Kuittinen and Jorma Sajaniemi. Teaching roles of
variables in elementary programming courses. In ITiCSE 9,
pages 57–61, 2004.
[LA03] Mark Lutz and David Ascher. Learning Python. O’Reilly &
Associates, Inc, Sebastopol, CA, 2003.
[Lak76] Imre Lakatos. Proofs and Refutations. Cambridge University
Press, Cambridge, United Kingdom, 1976.
[McC04] Steve McConnell. Code Complete: A Practical Handbook
of Software Construction. Microsoft Press, Redmond, WA,
2004.
[Pyt] Python education special interest group (edu-sig).
http://www.python.org/community/sigs/current/edu-sig/.
[VH07] John L. Viescas and Michael J. Hernandez. SQL Queries
for Mere Mortals: A Hands-On Guide to Data Manipulation in
SQL. Addison-Wesley, Reading, MA, 2007.
[Whi03] James A. Whittaker. How to Break Software. Addison-Wes-
ley, Reading, MA, 2003.
[Wil05] Greg Wilson. Data Crunching: Solve Everyday Problems
using Java, Python, and More. The Pragmatic Programmers,
LLC, Raleigh, NC, and Dallas, TX, 2005.
[Win06] Jeannette M. Wing. Computational thinking. Communica-
tions of the ACM, 49(3):33–35, 2006.
[WT04] James A. Whittaker and Herbert H. Thompson. How to
Break Software Security. Addison-Wesley, Reading, MA,
2004.
[Zel03] John Zelle. Python Programming: An Introduction to Com-
puter Science. Franklin Beedle & Associates, Wilsonville,
OR, 2003.
Report erratum
this copy is (P1.0 printing, April 2009)Prepared exclusively for Trieu Nguyen
IndexSymbols< (less than) operator, 111, 330
<= (less than or equal to) operator, 111,
330
with strings, 117
> (greater than) operator, 111, 330
with strings, 117
>= (greater than or equal to) operator,
111, 330
* (asterisk)
multiplication operator, 22, 25
working with lists, 89
selecting all database columns, 329
string repeat operator, 42
for variable parameter lists, 240, 261
** (exponentiation) operator, 22, 25
**= (exponentiation assignment)
operator, 29
*= (multiplication assignment) operator,
29
\ (backslash) as escape character, 43
[ ] (brackets)
for dictionary values, 191
empty list ([ ]), 84
for lists, 81
{ } (curly brackets) for dictionaries, 192
: (colon)
in functions, 30
in list slices, 93
, (comma), ending print statements, 46
= (assignment) operator, 26, 27, 111
= (equal to) operator (databases), 330
== (equal to) operator, 111
with objects, 277
!= (not equal to) operator, 111, 330
- (negation) operator, 22, 25, 36
- (subtraction) operator, 22, 25
with objects, 277
-= (subtraction assignment) operator,
29
( ) (parentheses) for tuples, 99
% (modulo) operator, 21, 22, 25
% (string formatting) operator, 46
%= (modulo assignment) operator, 29
. (period) for identifying modules, 51
+ (addition) operator, 22, 25
with lists, 88, 97
with objects, 277
+ (concatenation) operator, 40
+ (unary plus) operator, 36
+= (addition assignment) operator, 29
# for comments, 104
" (quotes)
for strings, 40
inside strings, 42
" (quotes)
for strings, 40
inside strings, 42
/ (division) operator, 20–22, 25
/= (division assignment) operator, 29
>>> prompt, 19
Aabs function, 33
Absolute value, 33
Accumulators, 258
Actual values, comparing with expected
values, 73
__add__ method, 277
add method (sets), 186
add_text function (media module), 66
Adding data to files, 181
Adding lists together, 88, 97
Addition (+) operator, 22, 25
with lists, 88, 97
with objects, 277
Addition assignment (+=) operator, 29
Prepared exclusively for Trieu Nguyen
AGGREGATION 354 CHARACTERS
Aggregation, database, 341
Algorithms, 203–212
defined, 203
measuring running time, 211
binary search, 221
categorizing by, 234
linear search, 216–218
sorting, 226–228
searching, 204–210
sentinel search, 216
using for loops, 215
sorting, 222–228
bubble sort, 235
insertion sort, 225–226
mergesort algorithm, 229–233
more-efficient examples, 228–229
selection sort, 222–226
Aliasing lists, 94–95
AND operator, 109
in database query conditions, 330
with numbers, 115–116
APIs (application programming
interfaces), 322
append method, 95
Append mode (opening files), 182
Application programming interfaces
(APIs), 322
Arguments, function, 31
Arithmetic operators, 25
combining with other operators,
112–114
ASC keyword (SELECT statements), 329
Ascending-order sorts, databases, 329
Assertions (in testing), 268
Assigning to variables, see Variables
Assignment statements, 26, 27, 29,
111
assigning list values, 85
multivalued assignment, 136
Associative operations, defined, 341
Asterisk (*)
exponentiation (**) operator, 22, 25
exponentiation assignment (**=)
operator, 29
multiplication assignment (*=)
operator, 29
multiplication operator, 22, 25
working with lists, 89
selecting all database columns, 329
string repeat operator, 42
for variable parameter lists, 240, 261
AVG function, 341
BBackground color, widget (Tkinter), 308
Backlash (\) as escape character, 43
Balanced colors, 200
Binary search, 218–221
built-in (bisect module), 221
running time, 221
bisect module, 221
bisect_left method (bisect module), 221
bisect_right method (bisect module), 221
Black-box testing, 249
BLOB data type (SQLite), 325, 326
bool type, 109, 270
Boolean logic, 108–118
one-way flags, 259
Boolean operators, 109–110
applying to integers, floats, and
strings, 114–118
combining with other operators,
112–114
for database query conditions, 330
BooleanVar type (Tkinter), 298
Border styles, Frame widget (Tkinter),
300
borderwidth attribute (Frame widget), 300
Boundary cases, 75, 252
Brackets [ ]
for dictionary values, 191
empty list ([ ]), 84
for lists, 81
Brackets { } for dictionaries, 192
break statement, 149–150, 153
Breakpoints, 144
Bubble sort, 235
Bugs, see Debugging
Built-in functions, 33, 55
for lists, 87–89
__builtins__ module, 55
Button widget (Tkinter), 295
Ccalendar module, 78
Call stack, 146
Canvas widget (Tkinter), 295
capitalize method, 61
Catching exceptions, 242, 248
Chaining methods, 62
Characters, 39
see also Strings
Prepared exclusively for Trieu Nguyen
CHECKBUTTON WIDGET (TKINTER) 355 DATA FILES
Checkbutton widget (Tkinter), 295, 314
Child classes, 286
Choices, 108–126
Boolean logic, 108–118
conditional statements in loops, 153
if statements, 118–124
nested, 124
storing conditionals, 125–126
choose_file function, 64, 67
class keyword, 271
Classes, 270
Color class (example), 271–275
constructors, 274–275
encapsulation, 280
inheritance, 285–287
methods of, 272–274
Clause, defining, 121
clear method (dictionaries), 194
clear method (sets), 186
click function, 302
Closing files, 267
Coding style, see Programming style,
notes on
Collections, see Dictionaries; Lists;
Sets; Tuples
Collisions, 198
Colon (:)
in functions, 30
in list slices, 93
Color class (example), 271–275
Colors, GUI, 308–310
Colors, image, 68–69
balanced, 200
common values, list of, 68
functions for, 68
grayscale, 158
models and representations of, 72
Columns in data records, 171–174
Columns in database tables, 321
Columns in printed output, 45
Combined operators, 28
Combining database tables, 334–339
self-joins, 342–344
Comma (,), ending print statements, 46
Command-line arguments, 103
filenames as, 161
Comment (#) character, 104
Comments in program code, 103
in modules, 59
commit method, 327
Comparison (==) operator, 111
with objects, 277
Comparison (=) operator (databases),
330
Comparisons, see Boolean logic;
Relational operators
Computational thinking, 14
Computer programs
defined, 13
executing, 17
statements of, 17
value of, 11
Concatenating lists, 88, 97
Concatenating strings, 40
Concatenation (+) operator, 40
Conditional logic, see Choices
Conditional loops, see while loops
Conditional statements in loops, 153
Conditionals, storing, 125–126
Conditions
for database queries, 329–330
defined, 118
connect method, 324
Constants (fixed values), 53, 256
CONSTRAINT keyword, 340
Constraints, database, 339–340
Constructors, 274–275
Containers, 258
continue statement, 151–153
Control flow statements, 108
see also for loops; if statements
Controllers, 302, 316
Conversion types, 46
COUNT function, 341
Counted loops, 131
for loops, 89–140
looping over multiple objects, 138
while loops, 140–147
Counters, 256
CREATE TABLE statement (SQL), 325
create_color function (media module), 68
crop_picture function (media module), 65
Curly brackets { } for dictionaries, 192
Current working directory, 66
Cursor (databases), 324
D%d conversion specifier, 46
darken function (media module), 68
Data collections, see Dictionaries;
Lists; Sets; Tuples
Data files, see Files
Prepared exclusively for Trieu Nguyen
DATA RECORDS IN FILES 356 EXCEPTION OBJECTS
Data records in files, 159
containing one value only, 160–170
extending across multiple lines,
177–179
with fields at fixed positions,
174–176
looking ahead when processing,
179–180
missing values in, 165
with multiple fields, 171–174
Data structure testing, 252
Data types for SQLite, 325, 327
Database management system (DBMS),
322
Databases, 321–346
aggregation, 341
connecting to, 324
creating tables, 325
grouping, 341
inserting data into, 326
joining tables, 334–339
self-joins, 342–344
keys and constraints, 339–340
missing data, NULL for, 333–334
nested queries, 345–346
saving changes to, 327
DBMS (database management system),
322
Debugging, 254–255
infinite loops, 144–146
see also Testing
def keyword, 30
Default parameter values (functions),
237
DELETE commands (SQL), 331
Deleting from databases, 330–331
Delimiters, 159
Depreciated, defined, 247
DESC keyword (SELECT statements), 329
Descending-order sorts, databases, 329
__dict__ variable, 280
Dictionaries, 190–197
classes as, 272
of dictionaries, 200
inverting, 197
looping over contents of, 193
methods for, 194
removing entries from, 193
updating, 192
difference method (sets), 186
dir function, 278
Directories
current working directory, 66
see also Files
distance function (media module), 68
Distinct collections, defined, 185
DISTINCT keyword (SELECT commands),
339
Division (mathematical), 21
Division (/) operator, 20–22, 25
Division assignment (/=) operator, 29
Docstrings, 103
in modules, 59
Dot (.) for identifying modules, 51
Double quotes, for strings, 40
Double quotes, inside strings, 42
DoubleVar type (Tkinter), 298
Downscaling images, 157
DROP commands (SQL), 331
Eelif keyword, 121
else clause, 123
Empty lists, 84
Empty string, 40
Encapsulation, 280
End-of-line (\n) character, 43, 44
Entry widget (Tkinter), 295, 300
enumerate function, 135–137
EOL errors, 44
__eq__ method, 277
Equal to (==) operator, 111
with objects, 277
Equal to (=) operator (databases), 330
Equality operators, 111
with objects, 277
Equality operators (databases), 330
Error management, see Exceptions
Error testing, see Debugging; Testing
Errors
indentation errors, 120
infinite loops, 142–146
out-of-range indices, 83
specific exceptions for, 244
variable assignment, 29
Errors of type, see Type errors
Escape characters, 42–43
Evaluating expressions, 19
see also Expressions
Event-driven programming, 294
except keyword, 242–243
Exception objects, 243–244
Prepared exclusively for Trieu Nguyen
EXCEPTIONS 357 GET_WIDTH FUNCTION (MEDIA MODULE)
Exceptions, 242–248
functions and, 245–246
raising, 242, 247
style tips, 247–248
try and except keywords, 242–243
Exclusive or, 110
Executing imported modules, 56
Executing programs, 17
Expected values, comparing with
actual values, 73
Exponentiation (**) operator, 22, 25
Exponentiation assignment (**=)
operator, 29
Exponentiation function (pow), 34
Expressions, 19–22
Extracting data from files, see Files,
reading
F%f conversion specifier, 46
False value, 108
converting to number, 118
fetchall method, 328
fetchone method, 328
Fields in data records
at fixed positions, 174–176
missing values, 165
multiline records, 177–179
multiple per record, 171–174
File objects, 101
Filenames as parameters, 161
Files, 159–183
closing, 267
creating new, 182
headers in, 162
as lists, 100–103
missing values, 165
for storing databases, 322
writing to, 181–182
see also Directories
Files, reading
looking ahead when processing,
179–180
multiline records, 177–179
multiple fields per record, 171–174
one field per record, 160–170
positional data, 174–176
find method, 61, 247
Finite numeric precision, 23, 24
Fixed values, 256
Fixed-width file formats, 174
FIXME comments, 104
Fixtures (in tests), 251
FLAT value, relief attribute, 300
float function, 47
float type, 21
Floating-point numbers, 21
applying Boolean operators to,
114–118
comparing integers to, 112
rounding, 34
user input of, 47
Folders, see Directories
Fonts, GUI, 307
for loops, 89–92, 131, 140
for linear search, 215
looping over dictionaries, 193
Foreground color, widget (Tkinter), 308
Format strings, 45
Formatted printing, 45
Frame widget (Tkinter), 295, 299
Frames, 146
Frozen sets, 190
Function calls, 31, 32
Function groups, see Modules
Functional testing, defined, 260
Functions, 30–34, 237–242
aliasing in, 94
built-in, 33, 55
default parameter values, 237–239
defining vs. calling, 32
exceptions and, 245–246
lambda functions, 303–307
for lists, 87–89
named parameters, 241
one-way flags, 259
testing with modules, 70–76
test_ functions, 71
using classes as, 271
variable parameter lists, 239–241
Functions in classes, see Methods
GGatherers, 258
General test cases, 252
get method (dictionaries), 194, 196
get_blue function (media module), 68
get_color function (media module), 68
get_green function (media module), 68
get_height function (media module), 65
get_red function (media module), 68
get_width function (media module), 65
Prepared exclusively for Trieu Nguyen
GLASS-BOX TESTING 358 ISSUBSET METHOD (SETS)
Glass-box testing, 249
Global variables, 303
Graphical user interfaces (GUIs),
294–318
basic construction of, 296–301
models, views, and controllers,
301–316
object-oriented, 316
styles of, 307–311
colors, 308–310
fonts, 307
layouts, 310–311
Tkinter module, 295
Graphics, see Images
Grayscale images, 158
Greater than (>) operator, 111, 330
with strings, 117
Greater than or equal to (>=) operator,
111, 330
grid layout manager, 311
GROUP BY clause (SELECT command), 342
Grouping, database, 341
GROOVE value, relief attribute, 300
GUI, see Graphical user interfaces
HHash codes, 189
Hash tables, 189–190
Header data in files, 162
Help documentation
comments in program code, 103
docstrings, 103
docstrings in modules, 59
help function, 51, 278
Helper functions, 169
Heterogeneity of lists, 85
Hexadecimal color representation, 72
IIDE, see Integrated development
environment
IDLE, 19
if statements, 118–124
nested, 124
using elif keyword, 121
using else clause, 123
using multiple, 121
Images, 64–67
pixels and colors, 68–69
scaling, 157
superimposing (example), 138
transforming or reflecting, 157
Immutability of numbers and strings,
86
import statement, 50, 53, 295n
Importing modules, 50–54
testing if module is imported, 58
Inclusive or, 110
Indentation in if statements, 119, 120
Independence of tests, 250
index method, 204
Indices for list items, 83
determining, 132–135
enumerating through list elements,
135–137
see also Lists
Infinite loops, 142–146
__init__ method, 275
Inner joins, 336, 342
Input from users, 46
loops for, 148–149
INSERT commands (SQL), 326
insert method, 95
Insertion sort, 225–226
insort_left method (bisect module), 221
insort_right method (bisect module), 221
Instance variables, 272
Instances of classes, 270
int type, 20, 270
operations with float type, 22
int function, 47
INTEGER data type (SQLite), 325
Integers, 20
applying Boolean operators to,
114–118
comparing to floating-point
numbers, 112
inserting in printed content, 46
user input of, 47
Integrated development environment
(IDE), 18
Interfaces, see Graphical user
interfaces (GUIs)
Internet, processing files over, 160
Interpreter, 18
Interrupting (breaking) loops, 149–150,
153
intersection method (sets), 186
IntVar type (Tkinter), 298
Inverting dictionaries, 197
islower method, 61
issubset method (sets), 186
Prepared exclusively for Trieu Nguyen
ISSUPERSET METHOD (SETS) 359 MERGESORT ALGORITHM
issuperset method (sets), 186
isupper method, 61
items method (dictionaries), 194
Iterations of loops
defined, 89
interrupting (breaking), 149–150,
153
skipping to next, 151–153
iteritems method (dictionaries), 195
JJoining database tables, 334–339
self-joins, 342–344
Joining strings, 40
KKey-value pairs, see Dictionaries
keys method (dictionaries), 194
Keys, database, 339–340
LLabel widget (Tkinter), 295
pack method, 296
Lambda functions, 303–307
Layouts, GUI, 310–311
len method, 87, 134
Less than (<) operator, 111, 330
Less than or equal to (<=) operator,
111, 330
with strings, 117
ligthen function (media module), 68
Linear search, 214–218
running time, 217–218
sentinel search, 216
using for loops, 215
List methods, 95–97
List searches, see Searching
List sorts, see Sorting
Listbox widget (Tkinter), 295
Lists, 81–105
aliasing, 94–95
alternatives to, 99–100
built-in functions for, 87–89
as containers, 258
disappearing (or so it seems), 96
empty lists, 84
enumerating elements of, 135–137
files as, 100–103
merging two sorted lists, 230–233
modifying, 85–86
nesting, 97–98
processing list items, 89–92
ragged lists, 139–140
slicing, 92–93
walking through, 208
load_picture function, 64
Local variables, 32–33, 146
Logic, conditional, see Choices
Logical operations with NULL, 334
Logical operators, see entries at
Boolean
Look and feel, GUI, 307–311
colors, 308–310
fonts, 307
layouts, 310–311
Loops, 131–154
counted loops, 131
looping over multiple objects, 138
over dictionary contents, 193
enumerating through list elements,
135–137
for loops, 89–92, 92, 131, 140
infinite loops, 142–146
interrupting with break, 149–150,
153
nesting, 90, 137
using break with, 150
skipping interactions with break,
151–153
style notes, 153
for user input, 148–149
walking through lists, 208
while loops, 140–147
see also specific loop by name
lower method, 61
M__main__ variable, 57
Maps, see Dictionaries
Mathematical expressions, 19
MAX function, 341
max method, 87
media module, 60, 64
pixel and color functions, 68
Memory model
lists, 83
for variable assignment, 26
Menu widget (Tkinter), 295, 314
Menubutton widget (Tkinter), 295
merge function, 231
Mergesort algorithm, 229–233
Prepared exclusively for Trieu Nguyen
MESSAGE WIDGET (TKINTER) 360 OPENING FILES
Message widget (Tkinter), 295
Methods, 61, 272–274
chaining together, 62
constructors, 274–275
for dictionaries, 194
inheritance, 285–287
for lists, 95–97
overriding, 287
polymorphism, 282–284
for sets, 185–187
special methods, 276–278
Methods for strings, see String
methods
MIN function, 341
min method, 87
Missing data in database, NULL for,
333–334
Missing values in data records, 165
Models, 302, 316
Modes for opening files, 181
Modes of file access, 100
Modules, 50–78
defining custom, 54–60
executing when imported, 56
for file processing, 159
for images (example), 64–67
importing, 50–54
style notes on, 76
for testing functions, 70–76
testing if imported, 58
writing comments in, 59
Modulo (%) operator, 21, 22, 25
Modulo assignment (%=) operator, 29
Mosaic filter, 157
Most-recent holder, 257
Most-wanted holder, 257
Multiline data records, 177–179
Multiline strings, 43–44
Multiplication (*) operator, 22, 25
with lists, 89
Multiplication assignment (*) operator,
29
Multivalued assignment, 136
Mutability of lists, 86, 94
Mutable variables (Tkinter), 297–299
N\n (newline) character, 43, 44
__name__ variable, 57
Named parameters, 241
with widgets (Tkinter), 297
Naming variables, 34
Negation (-) operator, 22, 25, 36
Nested if statements, 124
Nested database queries, 345–346
Nested lists, 97–98
ragged lists, 139–140
Nested loops, 90, 137
using break with, 150
Newline (\n) character, 43, 44
None value, returned from list
methods, 96
Normalizing strings, 44
Nose library, 71
Not equal to (!=) operator, 111, 330
NOT NULL keywords, 333
NOT operator, 109
in database query conditions, 330
with numbers, 114
NULL data type, 325
how to use, 333
for missing data, 333, 334
Number comparisons, see Relational
operators
Numbers
converting True and False to, 118
as immutable, 86
testing with, 252
from user input, 47
Numeric precision, 23, 24
Numerical analysis, 24
OObject-oriented GUIs, 316
Object-oriented programming (OOP),
270–293
Color class (example), 271–275
special methods, 276–278
theory of, 280–287
encapsulation, 280
inheritance, 285–287
polymorphism, 282–284
Objects, 63, 270
constructors, 274–275
encapsulation, 280
exception objects, 243–244
inheritance, 285–287
polymorphism, 282–284
One-way flags, 259
OOP, see Object-oriented programming
open function, 100
Opening files, 162, 181
Prepared exclusively for Trieu Nguyen
OPERANDS 361 RAW_INPUT FUNCTION
see also Reading files; Writing to files
Operands, 19
Operating system (OS), 17
Operations in classes, see Methods
Operations in computer programs, 13
defining new, 14
Operations, database, see
Transactions, database
Operators, 19
arithmetic operators, 25
Boolean operators, 109–110
applying to integers, floats, and
strings, 114–118
combined operators, 28
combining different types of,
112–114
Boolean operators
for database query conditions, 330
relational operators
for database query conditions, 330
overloaded, defined, 40
overloading, 278
relational operators, 111–112
with set methods, 187
OR operator, 109
in database query conditions, 330
NULL with, 334
with numbers, 115–116
ORDER BY clause, 328
Order of precedence, see Operator
precedence
Out-of-range indices, 83
Outdated comments, 105
Overloaded operators, defined, 40
Overloading operators, 278
Overriding methods, 287
Ppack method, entry widget, 310
Parameters, function, 31, 33, 146
default values for, 237–239
filenames as, 161
named, 241
self parameter, 273
variable parameter lists, 239–241
Parent classes, 286
Parent widgets (Tkinter), 296
Parentheses ( ) for tuples, 99
pack method (Label widget), 296
pass keyword, 271
pass statement, 72
Patterns, 256–260
containers, 258
fixed values, 256
gatherers (accumulators), 258
most-recent holders, 257
most-wanted holders, 257
one-way flags, 259
steppers and counters, 256
temporary variables, 258
Period (.) for identifying modules, 51
Pictures, see Images
Pixels, 68–69
Polymorphism, 282–284
pop method, 95, 199
Positional data in files, 174–176
pow function, 34
Precedence, operator, 25, 112–114
Precision, numeric, 23, 24
PRIMARY KEY clause (CONSTRAINT
command), 339
print statement, 44–46
formatted printing, 45
Processing files, see Files
Profiling, 211
binary search algorithms, 221
families of algorithms, 234
linear search algorithms, 216–218
sort algorithms, 226–228
Programming style, see Style,
programming
Programs, see Computer programs
Prompt, Python (>>>), 19
Python, good for teaching with, 15
QQuality assurance (QA), 70
Queries, database, 327–330
avoiding duplicate results, 339
conditions for, 329–330
nested, 345–346
Quotes, for strings, 40
Quotes, inside strings, 42
RRagged lists, 139–140
raise keyword, 247
RAISED value, relief attribute, 300
Raising exceptions, 242, 247
range function, 132–135
raw_input function, 46
in loops, 148–149
Prepared exclusively for Trieu Nguyen
READ METHOD 362 SELECT COMMANDS (SQL)
read method, 101
Read mode (opening files), 181
Reading files
looking ahead when, 179–180
with multiline records, 177–179
with multiple fields per record,
171–174
with one field per record, 170
with positional data, 174
Reading from databases, 327–330
avoiding duplicates, 339
query conditions, 329–330
Reading from files, 100
readline method, 101
REAL data type (SQLite), 325
Real numbers, precision of, 24
Records (data) in files, 159
extending across multiple lines,
177–179
looking ahead when processing,
179–180
missing values in, 165
with multiple fields, 171–174
one per line, 160–170
Records in database tables, 321
updating or deleting, 330–331
Reflecting images, 157
Regression testing, 250
Relational databases, 321–346
aggregation, 341
connecting to, 324
creating tables, 325
grouping, 341
inserting data into, 326
joining tables, 334–339
self-joins, 342–344
keys and constraints, 339–340
missing data, NULL for, 333–334
nested queries, 345–346
retrieving data from, 327–330
query conditions, 329–330
without duplicates, 339
saving changes to, 327
transactions, 331–333
updating and deleting from, 330–331
Relational operators, 111–112
combining with other operators,
112–114
for database query conditions, 330
with strings, 117
relief attribute (Frame widget), 300
Remainder (%) operator, 21, 22, 25
Remainder assignment (%=) operator,
29
Reminders, comments as, 104
remove method, 95, 186, 206
Repeating strings with * operator, 42
Repeating things, see Loops
replace method, 61
__repr__ method, 276
Resizing images, 157
Return address, 146
return statement
function calls, 30, 31
Return value, 147
reverse method, 95
RGB color model, 72
RIDGE value, relief attribute, 300
Rolling back transactions, 331
Root window (Tkinter), 295
adding widgets to, 296
round function, 34
Rounding numbers, 21, 34
Running time, measuring, 211
binary search, 221
categorizing algorithms by, 234
linear search, 216–218
sort algorithms, 226–228
Runtime stack, 146
S%s conversion specifier, 46
save_as function (media module), 65
Saving changes to databases, 327
Saving data to files, see Files, writing to
Scaling images, 157
Scope, variable, 33
Searching
binary search, 218–221
built-in (bisect module), 221
timing algorithms for, 221
choosing an algorithm for, 204–210
linear search, 214–218
sentinel search, 216
timing algorithms for, 216–218
using for loops, 215
test strategies for, 252
SELECT commands (SQL), 327–330
DISTINCT keyword, 339
GROUP BY clause, 342
nested, 345–346
ORDER BY clause, 328
Prepared exclusively for Trieu Nguyen
SELECTION SORT 363 SUM FUNCTION
query conditions, 329–330
Selection sort, 222–226
self parameter, 273
Self-joins, 342–344
Semantics, code, 278
Sentinel search, 216
Sequences, see Dictionaries; Lists;
Sets; Tuples
Sequences, enumerating elements of,
135–137
Set operations, 185–187
set_blue function (media module), 68
set_color function (media module), 68
set_green function (media module), 68
set_red function (media module), 68
Sets, 185–190
frozen sets, 190
storage in hash tables, 189–190
Shells, 18
show function, 64
side argument, pack method, 310
Simplest interesting test cases, 252
Single quotes, for strings, 40
Single quotes, inside strings, 42
Size, picture
determining, 65
editing, 65
Skipping iterations of loops, 151–153
Slicing lists, 92–93
sort method, 95
sorted function, 197
Sorting, 222–228
bubble sort, 235
in databases, 328
improving efficiency of, 228–229
insertion sort, 225–226
mergesort algorithm, 229–233
selection sort, 222–226
timing algorithms for, 226–228
Sparse vectors, 201
split method, 61, 169
SQL (Structured Query Language), 323
SQLite library, 322, 323
data types for, 325, 327
sqrt function, 51
Square brackets [ ]
for dictionary values, 191
empty list ([ ]), 84
for lists, 81
Stack Data tab (Wing 101), 145
Starred parameters (functions), 239
startswith method, 61
Statements
see also Expressions
Statements, program, 17
Step Into button (Wing 101), 145
Step Over button (Wing 101), 145
Step size, 133
Stepper variables, 256
Stopping (breaking) loops, 149–150,
153
__str__ method, 276
String formatting (%) operator, 46
String methods, 61–62
Strings, 39–47
applying Boolean operators to,
114–118
as containers, 258
escape characters, 42–43
as immutable, 86
inserting in printed content, 46
multiline, 43–44
normalizing, 44
print statement, 44–46
as sequences, 99
stripping whitespace from, 102
user input, 46
StringVar type (Tkinter), 298
strip method, 61, 102
Structure testing, 252
Structured Query Language (SQL), 323
Style, programming
being too clever, 117
debugging strategies, 255
exceptions, 247–248
loops, 153
naming variables, 34
test-driven development, 253
using modules, 76
writing comments, 104
Styles of graphical user interfaces
(GUIs), 307–311
colors, 308–310
fonts, 307
layouts, 310–311
__sub__ method, 277
Subclasses, 286
Subtraction (-) operator, 22, 25
with objects, 277
Subtraction assignment (-=) operator,
29
SUM function, 341
Prepared exclusively for Trieu Nguyen
SUM METHOD 364 VARIABLES
sum method, 87
SUNKEN value, relief attribute, 300
Superclasses, 286
Superimposing images (example), 138
symmetric_difference method (sets), 186
Syntax errors, 30
sys.argv variable, 103
filenames as parameters, 161
T\t (tab) character, 43, 45
Tab (\t) character, 43, 45
Tables in relational databases, 321
combining (joining), 334–339
self-joins, 342–344
creating, 325
removing, 331
TDD (test-driven development), 253
Temporaries, 258
test_ files, 71
Test-driven development (TDD), 253
Testing, 249–254
black-box and glass-box testing, 249
cases for, types of, 252
disappearing lists, 96
functional testing, defined, 260
functions, with modules, 70–76
independence of tests, 250
limitations of, 251
regression testing, 250
terminology of, 249–250
unit testing, 250
using reminder comments, 104
see also Debugging
text attribute (Label widget), 296
TEXT data type (SQLite), 325
text_factory method, 328
Text strings, see Strings
Text style, GUI
fonts of, 307
Text widget (Tkinter), 295, 312
textvariable attribute (Label widget), 298
Three-valued logic, 334
Tkinter module, 295
mutable variables, 297–299
TODO comments, 104
Top-down design, 203
TopLevel widget (Tkinter), 295
Transactions, database, 331–333
Transforming images, 157
Troubleshooting, see Debugging;
Testing
comments about bugs, 104
True value, 108
converting to number, 118
Truth tables, 110
try keyword, 242–243
Tuples, 99, 190
enumerating elements of, 135–137
Type errors
adding different data types, 41
adding lists and strings, 88
wrong number of parameters, 33
Types, 20, 22–25, 270
creating new, 270
see also specific type by name
UUnary minus (-) operator, 22, 25, 36
Unary plus (+) operator, 36
Undoing transactions, 331
Unicode scheme, 326
union method (sets), 186
Unique IDs in databases, 339
Unit testing, 250
defined, 260
independence of tests, 250
limitations of, 251
Unordered collections, defined, 185
UPDATE commands (SQL), 330
update method (dictionaries), 194
Updating databases, 330–331
Updating dictionaries, 192
upper method, 61
urllib module, 160
urlopen function (urllib module), 160
User input, 46
loops for, 148–149
User interfaces, see Graphical user
interfaces (GUIs)
VValues in data records, see Fields in
data records
values method (dictionaries), 194
Variable parameter lists, 239–241
Variables, 25–30
assigning lists to, 83
assigning values to, 25, 26, 29
multivalued assignment, 136
conditionals stored as, 125–126
Prepared exclusively for Trieu Nguyen
VIEWS 365 WRITING TO FILES
creating, 25
global, 303
importing from modules, 52
instance variables, 272
local, 32–33, 146
most-recent holders, 257
most-wanted holders, 257
mutable, Tkinter, 297–299
naming, tips for, 34
scopeof, 33
steppers, 256
temporaries, 258
Views, 301, 316
Virtual machine, 18
WWalking through lists, 208
WHERE keyword (SELECT statements), 329
while loops, 140–147
Whitespace, stripping from strings, 102
Widgets (Tkinter), 295
adding to root window, 296
colors of, 308
layout of, 310
parent widgets, 296
Windowing systems, see Graphical
user interfaces (GUIs)
Wing 101, 18, 144
Working directory, 66
Write mode (opening files), 181
Writing to files, 181–182
Prepared exclusively for Trieu Nguyen
The Pragmatic BookshelfAvailable in paperback and DRM-free PDF, our titles are here to help you stay on top of
your game. The following are in print as of April 2009; be sure to check our website at
pragprog.com for newer titles.
Title Year ISBN Pages
Advanced Rails Recipes: 84 New Ways to Build
Stunning Rails Apps
2008 9780978739225 464
Agile Retrospectives: Making Good Teams Great 2006 9780977616640 200
Agile Web Development with Rails: Second
Edition
2006 9780977616633 719
Agile Web Development with Rails, Third Edition 2009 9781934356166 784
Augmented Reality: A Practical Guide 2008 9781934356036 328
Behind Closed Doors: Secrets of Great
Management
2005 9780976694021 192
Best of Ruby Quiz 2006 9780976694076 304
Core Animation for Mac OS X and the iPhone:
Creating Compelling Dynamic User Interfaces
2008 9781934356104 200
Data Crunching: Solve Everyday Problems
using Java, Python, and More
2005 9780974514079 208
Deploying Rails Applications: A Step-by-Step
Guide
2008 9780978739201 280
Design Accessible Web Sites: 36 Keys to
Creating Content for All Audiences and
Platforms
2007 9781934356029 336
Desktop GIS: Mapping the Planet with Open
Source Tools
2008 9781934356067 368
Developing Facebook Platform Applications with
Rails
2008 9781934356128 200
Enterprise Integration with Ruby 2006 9780976694069 360
Enterprise Recipes with Ruby and Rails 2008 9781934356234 416
Everyday Scripting with Ruby: for Teams,
Testers, and You
2007 9780977616619 320
FXRuby: Create Lean and Mean GUIs with Ruby 2008 9781934356074 240
From Java To Ruby: Things Every Manager
Should Know
2006 9780976694090 160
GIS for Web Developers: Adding Where to Your
Web Applications
2007 9780974514093 275
Google Maps API, V2: Adding Where to Your
Applications
2006 PDF-Only 83
Groovy Recipes: Greasing the Wheels of Java 2008 9780978739294 264
Hello, Android: Introducing Google’s Mobile
Development Platform
2008 9781934356173 200
Interface Oriented Design 2006 9780976694052 240
Continued on next page
Prepared exclusively for Trieu Nguyen
WRITING TO FILES 367 WRITING TO FILES
Title Year ISBN Pages
Learn to Program, 2nd Edition 2009 9781934356364 230
Manage It! Your Guide to Modern Pragmatic
Project Management
2007 9780978739249 360
Mastering Dojo: JavaScript and Ajax Tools for
Great Web Experiences
2008 9781934356111 568
My Job Went to India: 52 Ways to Save Your Job 2005 9780976694014 208
No Fluff Just Stuff 2006 Anthology 2006 9780977616664 240
No Fluff Just Stuff 2007 Anthology 2007 9780978739287 320
Practices of an Agile Developer 2006 9780974514086 208
Pragmatic Project Automation: How to Build,
Deploy, and Monitor Java Applications
2004 9780974514031 176
Pragmatic Thinking and Learning: Refactor Your
Wetware
2008 9781934356050 288
Pragmatic Unit Testing in C# with NUnit 2007 9780977616671 176
Pragmatic Unit Testing in Java with JUnit 2003 9780974514017 160
Pragmatic Version Control Using Git 2008 9781934356159 200
Pragmatic Version Control using CVS 2003 9780974514000 176
Pragmatic Version Control using Subversion 2006 9780977616657 248
Programming Erlang: Software for a Concurrent
World
2007 9781934356005 536
Programming Groovy: Dynamic Productivity for
the Java Developer
2008 9781934356098 320
Programming Ruby: The Pragmatic
Programmers’ Guide, Second Edition
2004 9780974514055 864
Prototype and script.aculo.us: You Never Knew
JavaScript Could Do This!
2007 9781934356012 448
Rails Recipes 2006 9780977616602 350
Rails for .NET Developers 2008 9781934356203 300
Rails for Java Developers 2007 9780977616695 336
Rails for PHP Developers 2008 9781934356043 432
Rapid GUI Development with QtRuby 2005 PDF-Only 83
Release It! Design and Deploy Production-Ready
Software
2007 9780978739218 368
Scripted GUI Testing with Ruby 2008 9781934356180 192
Ship it! A Practical Guide to Successful Software
Projects
2005 9780974514048 224
Stripes ...And Java Web Development Is Fun
Again
2008 9781934356210 375
TextMate: Power Editing for the Mac 2007 9780978739232 208
The Definitive ANTLR Reference: Building
Domain-Specific Languages
2007 9780978739256 384
The Passionate Programmer: Creating a
Remarkable Career in Software Development
2009 9781934356340 200
Continued on next page
Prepared exclusively for Trieu Nguyen
Title Year ISBN Pages
ThoughtWorks Anthology 2008 9781934356142 240
Ubuntu Kung Fu: Tips, Tricks, Hints, and Hacks 2008 9781934356227 400
Prepared exclusively for Trieu Nguyen
The Pragmatic BookshelfThe Pragmatic Bookshelf features books written by developers for developers. The titles
continue the well-known Pragmatic Programmer style and continue to garner awards and
rave reviews. As development gets more and more difficult, the Pragmatic Programmers
will be there with more titles and products to help you stay on top of your game.
Visit Us OnlinePractical Programming’s Home Page
http://pragprog.com/titles/gwpy
Source code from this book, errata, and other resources. Come give us feedback, too!
Register for Updates
http://pragprog.com/updates
Be notified when updates and new books become available.
Join the Community
http://pragprog.com/community
Read our weblogs, join our online discussions, participate in our mailing list, interact
with our wiki, and benefit from the experience of other Pragmatic Programmers.
New and Noteworthy
http://pragprog.com/news
Check out the latest pragmatic developments, new titles and other offerings.
Buy the BookIf you liked this eBook, perhaps you’d like to have a paper copy of the book. It’s available
for purchase at our store: pragprog.com/titles/gwpy.
Contact UsOnline Orders: www.pragprog.com/catalog
Customer Service: [email protected]
Non-English Versions: [email protected]
Pragmatic Teaching: [email protected]
Author Proposals: [email protected]
Contact us: 1-800-699-PROG (+1 919 847 3884)
Prepared exclusively for Trieu Nguyen